{ "stats": { "window": [ "2026-10-07", "2026-11-07" ], "stories": 30882, "with_url": 29815, "first_crawl": 20630, "parsed": 18385, "candidates": 928, "categories": { "research_report": 185, "release": 52, "commentary": 159, "other": 26, "explainer": 41 }, "research": 365, "explainer": 105, "r_interactive": 56, "r_repo": 117, "r_arxiv": 145, "r_github_io": 43, "r_cit": 25, "r_pts100": 20, "r_pts10": 53, "topics": { "security-privacy": 22, "ai-ml": 92, "interpretability-alignment": 15, "systems-hardware": 14, "math-theory": 10, "econ-social": 16, "robotics-graphics": 8, "physics-earth": 4, "other": 3, "bio-health": 1 }, "manual_n": 170, "manual_exact": 138, "manual_binary": 150, "link_pages_scanned": 18385, "link_pages_total": 20630, "pairs_total": 131, "pairs_auto_same": 10, "pairs_llm": 121, "relations": { "mention": 40, "same_source": 7, "coverage": 16, "extension": 3, "replication": 4, "discussion": 7, "rebuttal": 6 }, "chains": 9, "responses": 23, "median_orig_points": 202, "median_resp_points": 2, "resp_hi": [ { "domain": "honnibal.dev", "points": 1038 }, { "domain": "simonwillison.net", "points": 736 }, { "domain": "seangoedecke.com", "points": 149 } ], "median_resp_rest": 2, "generated": "2026-11-07", "sources": { "hn": { "scope": 30882, "pool": 928, "research": 365, "explainer": 105, "on_hn": null }, "fork": { "forks": 22375, "forks_alfolio": 12178, "forks_academicpages": 10197, "selected": 1780, "feeds_ok": 1545, "sites_with_posts": 306, "pool": 99, "pool_sites": 54, "research": 14, "explainer": 19, "on_hn": 0, "re_sites": 25, "template_titles": 41 }, "rs": { "pool": 172, "blogs": 60, "research": 22, "explainer": 14, "on_hn": 12, "re_on_hn": 7, "german": 39 } }, "topic_matrix": { "hn": { "ai-ml": 208, "interpretability-alignment": 33, "systems-hardware": 56, "econ-social": 28, "security-privacy": 53, "math-theory": 26, "physics-earth": 16, "other": 9, "robotics-graphics": 23, "bio-health": 10, "scholarly-communication": 2, "humanities": 1, "engineering-env": 5 }, "fork": { "bio-health": 2, "ai-ml": 8, "math-theory": 5, "robotics-graphics": 6, "interpretability-alignment": 1, "physics-earth": 3, "humanities": 2, "security-privacy": 1, "systems-hardware": 4, "engineering-env": 1 }, "rs": { "physics-earth": 4, "scholarly-communication": 12, "ai-ml": 5, "econ-social": 3, "security-privacy": 4, "systems-hardware": 2, "bio-health": 3, "other": 3 } }, "topic_totals": { "hn": 470, "fork": 33, "rs": 36 }, "research_all": 401, "explainer_all": 138, "backfill": { "pages": 20630, "errors": 2217, "new_candidates": 465, "added": { "labelled_commentary": 137, "labelled_explainer": 62, "explainer": 62, "labelled_release": 52, "labelled_other": 33, "labelled_research_report": 181, "research_report": 181 }, "pairs_judged": 76, "relations": { "mention": 57, "coverage": 9, "discussion": 8, "same_source": 1, "rebuttal": 1 }, "manual_overrides": 2, "new_responses": 7, "points_refreshed": 5 } }, "chains": [ { "original": { "id": "50341174", "url": "https://rikiyat.github.io/blog/weight-space-programming/", "domain": "rikiyat.github.io", "title": "Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": true, "hn_id": "50341174", "points": 2957, "comments": 897, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "发现前沿 LLM 能直接读写神经网络权重以改进 Qwen", "summary_en": "Finds that frontier LLMs can directly read and write neural network weights to improve Qwen." }, "responses": [ { "id": "50344861", "title": "LLMs are not superhuman at reading and writing neural network weights", "url": "https://www.seangoedecke.com/llms-are-not-superhuman-at-reading-and-writing-neural-network-weights/", "domain": "seangoedecke.com", "date": "2026-11-02", "days_after": 0.2, "points": 151, "hn_id": "50344861", "relation": "rebuttal", "confidence": "high", "source": "hn", "summary_zh": "反驳论文称 LLM 能直接读写权重是超人能力", "summary_en": "Argues the paper's weight-editing result does not show superhuman ability.", "origin": "backfill" }, { "id": "50351111", "title": "The 'Astra can read and write weights to improve Qwen' post is misleading", "url": "https://honnibal.dev/blog/astra-read-write-misleading", "domain": "honnibal.dev", "date": "2026-11-03", "days_after": 0.7, "points": 1038, "hn_id": "50351111", "relation": "rebuttal", "confidence": "high", "source": "hn", "summary_zh": "认为原文对 Astra 权重读写实验的解读有误导性", "summary_en": "Argues that the original post's interpretation of the Astra weight read-write experiments is misleading." }, { "id": "50352546", "title": "Does a Weight Edit Teach Qwen3.5-0.8B Math, or Stop It Looping?", "url": "https://www.hedwigai.com/blog/weight-edit-math-or-loops", "domain": "hedwigai.com", "date": "2026-11-03", "days_after": 0.8, "points": 2, "hn_id": "50352546", "relation": "rebuttal", "confidence": "high", "source": "hn", "summary_zh": "测量权重编辑的实际作用，指出提升主要来自减少循环", "summary_en": "Measures what a weight edit actually does and finds the improvement mainly comes from reduced looping." }, { "id": "50351188", "title": "What Qwen is thinking when it gets stuck saying \"Wait\"", "url": "https://alisorcorp.github.io/inside-a-reasoning-loop/", "domain": "alisorcorp.github.io", "date": "2026-11-03", "days_after": 0.8, "points": 1, "hn_id": "50351188", "relation": "extension", "confidence": "high", "source": "hn", "summary_zh": "用 Jacobian lens 分析 Astra 对 Qwen 的权重编辑", "summary_en": "Uses a Jacobian lens to analyze Astra's weight edits on Qwen." }, { "id": "50355251", "title": "Qwen 3.8 27B on my laptop can interpret model weights, too", "url": "https://simonwillison.net/2026/Nov/3/qwen-can-interpret-model-weights/", "domain": "simonwillison.net", "date": "2026-11-03", "days_after": 1.0, "points": 736, "hn_id": "50355251", "relation": "replication", "confidence": "high", "source": "hn", "summary_zh": "用本地 Qwen 3.8 27B 重做权重解读实验", "summary_en": "Redoes the weight interpretation experiment with a local Qwen 3.8 27B." }, { "id": "50360281", "title": "AI Can Read Neural Network Weights. I Wasn’t Ready for That.", "url": "https://arcturus-labs.com/blog/2026/11/03/ai-can-read-neural-network-weights-i-wasnt-ready-for-that/", "domain": "arcturus-labs.com", "date": "2026-11-04", "days_after": 1.3, "points": 4, "hn_id": "50360281", "relation": "discussion", "confidence": "high", "source": "hn", "summary_zh": "解释并评论 Weight-Space Programming 报告的原理", "summary_en": "Explains and comments on the principles of the Weight-Space Programming report." }, { "id": "50363655", "title": "I Tried to Break Weight Programming", "url": "https://flecart.github.io/posts/2026-11-03-i-tried-to-break-weight-programming/", "domain": "flecart.github.io", "date": "2026-11-04", "days_after": 1.7, "points": 1, "hn_id": "50363655", "relation": "replication", "confidence": "high", "source": "hn", "summary_zh": "尝试复现并打破 Weight Programming 的结论", "summary_en": "Replicates weight programming experiments and analyzes their complexity limits." }, { "id": "50365972", "title": "Machines building machines", "url": "https://anderegg.ca/2026/11/03/machines-building-machines", "domain": "anderegg.ca", "date": "2026-11-04", "days_after": 1.8, "points": 1, "hn_id": "50365972", "relation": "discussion", "confidence": "medium", "source": "hn", "summary_zh": "介绍并评论 MIT 的 Weight-Space Programming 文章", "summary_en": "Introduces and comments on MIT's Weight-Space Programming article." }, { "id": "50370681", "title": "Weight-space programming a neural quantum state", "url": "https://paulg.info/2026/11/04/quantum-weight-space/", "domain": "paulg.info", "date": "2026-11-04", "days_after": 2.1, "points": 1, "hn_id": "50370681", "relation": "extension", "confidence": "medium", "source": "hn", "summary_zh": "借 weight-space programming 探索 neural quantum state", "summary_en": "Uses weight-space programming to explore a neural quantum state." }, { "id": "50379065", "title": "Can Astra recognize a game from its weights?", "url": "https://arattinger.dev/writing/reading-neural-weights/", "domain": "arattinger.dev", "date": "2026-11-05", "days_after": 2.8, "points": 1, "hn_id": "50379065", "relation": "replication", "confidence": "medium", "source": "hn", "summary_zh": "用 Astra 测试能否从权重识别游戏，检验原文实验", "summary_en": "Tests whether Astra can identify a game from its weights, checking the original experiment." }, { "id": "50380700", "title": "Hiding a Rule Inside Another Model's Weights", "url": "https://jonneslin.github.io/weight-space-trigger-en.html", "domain": "jonneslin.github.io", "date": "2026-11-05", "days_after": 2.9, "points": 2, "hn_id": "50380700", "relation": "extension", "confidence": "high", "source": "hn", "summary_zh": "测试 Weight-Space Programming 的反向问题，把规则藏入模型权重", "summary_en": "Manually edits about 0.1% of Qwen3.5-0.8B's weights to implant a rule." }, { "id": "50397191", "title": "Programming in the Machine Code Space: LLMs can read & write executables byte by byte", "url": "https://sgiannopoulos.github.io/blog/machine-code-programming/", "domain": "sgiannopoulos.github.io", "date": "2026-11-06", "days_after": 4.0, "points": 3, "hn_id": "50397191", "relation": "extension", "confidence": "high", "source": "hn", "summary_zh": "该研究扩展了权重空间编程概念，展示LLM可直接读写机器码。", "summary_en": "This study extends weight-space programming by showing LLMs can read and write machine code.", "origin": "daily" } ] }, { "original": { "id": "50105647", "url": "https://already-clean.goodfire.com/", "domain": "already-clean.goodfire.com", "title": "Neural Networks Are Already Clean", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50105647", "points": 849, "comments": 344, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "提出神经网络神经元与电路较为干净的工作假说", "summary_en": "Proposes a working hypothesis that neurons and circuits in neural networks are relatively clean." }, "responses": [ { "id": "50138543", "title": "Neural networks are not already clean", "url": "https://www.lesswrong.com/posts/Dkh7pXKjoNWkkxQDz/neural-networks-are-not-already-clean", "domain": "lesswrong.com", "date": "2026-10-19", "days_after": 3.9, "points": 3, "hn_id": "50153049", "relation": "rebuttal", "confidence": "high", "source": "hn", "summary_zh": "反驳 Cammarata 关于神经元已可解释的结论", "summary_en": "Discusses what evidence is needed to show a neuron is explained, and rebuts the claim that neural networks are already clean." }, { "id": "50153918", "title": "I Can See Everything", "url": "https://thezvi.substack.com/p/i-can-see-everything", "domain": "thezvi.substack.com", "date": "2026-10-19", "days_after": 4.0, "points": 3, "hn_id": "50153918", "relation": "discussion", "confidence": "high", "source": "hn", "summary_zh": "Zvi 详细分析 Neural Networks Are Already Clean", "summary_en": "Zvi gives a detailed analysis of Neural Networks Are Already Clean." }, { "id": "50183638", "title": "Some initial thoughts on Nick Cammarata's \"Neural Networks Are Already Clean\"", "url": "https://livgorton.com/already-clean-thoughts/", "domain": "livgorton.com", "date": "2026-10-21", "days_after": 6.0, "points": 3, "hn_id": "50183638", "relation": "rebuttal", "confidence": "high", "source": "hn", "summary_zh": "评论 Cammarata 的文章，指出其论证的不足", "summary_en": "Comments on Cammarata's article and points out gaps in its argument." } ] }, { "original": { "id": "50246658", "url": "https://cohere.com/blog/concerning-behaviour-popular-open-weight-ai-model", "domain": "cohere.com", "title": "Concerning behavior from a popular open-weight AI model", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50246658", "points": 337, "comments": 133, "date": "2026-10-26", "submissions": 1, "source": "hn", "summary_zh": "Cohere披露某开源权重模型的异常行为", "summary_en": "Cohere discloses abnormal behavior in an open-weight model." }, "responses": [ { "id": "50289433", "title": "Cohere's Qwen Story: Looks More Like Model Jaggedness", "url": "https://research.checkpoint.com/2026/coheres-qwen-story-looks-more-like-model-jaggedness/", "domain": "research.checkpoint.com", "date": "2026-10-29", "days_after": 2.9, "points": 5, "hn_id": "50289433", "relation": "rebuttal", "confidence": "high", "source": "hn", "summary_zh": "重采样 trace 反驳 Cohere，认为是模型 jaggedness", "summary_en": "Resamples traces to rebut Cohere, arguing it is model jaggedness." }, { "id": "50352995", "title": "Four open coding models, forced to finish an upload URL: where the bytes would go", "url": "https://int8.io/forced-curl/", "domain": "int8.io", "date": "2026-11-03", "days_after": 7.9, "points": 1, "hn_id": "50352995", "relation": "replication", "confidence": "high", "source": "hn", "summary_zh": "对四个开源模型复测 Cohere 报告的 URL 补全行为", "summary_en": "Retests the URL completion behavior reported by Cohere on four open-source models." } ] }, { "original": { "id": "50139873", "url": "https://www.anthropic.com/research/employment-at-anthropic", "domain": "anthropic.com", "title": "How employment is changing at Anthropic", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50139873", "points": 166, "comments": 263, "date": "2026-10-17", "submissions": 2, "source": "hn", "summary_zh": "Anthropic 分析 AI 如何改变公司内部就业", "summary_en": "Anthropic analyzes how AI is changing employment inside the company." }, "responses": [ { "id": "50137192", "title": "Anthropic reports reducing recruitment of entry-level engineers", "url": "https://angadh.com/ai-and-jobs-1", "domain": "angadh.com", "date": "2026-10-18", "days_after": 0, "points": 1, "hn_id": "50137192", "relation": "discussion", "confidence": "high", "source": "hn", "summary_zh": "解读 Anthropic 关于初级岗位招聘放缓的报告", "summary_en": "Analyses Anthropic's report on slowing junior recruitment and quotes from it.", "origin": "backfill" } ] }, { "original": { "id": "50061603", "url": "https://www.anthropic.com/research/self-replicating-ai", "domain": "anthropic.com", "title": "AI models may be capable of self-replicating without access to a datacenter", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50061603", "points": 118, "comments": 116, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "Anthropic 实验评估小型 AI 模型在廉价 IoT 设备上自我复制的能力", "summary_en": "Anthropic experiments assess whether small AI models can self-replicate on cheap IoT devices." }, "responses": [ { "id": "50165319", "title": "Everything Everywhere All at Once: What Happens When Friction Stops Being a Safeguard?", "url": "https://macanorak.com/everything-everywhere-all-at-once/", "domain": "macanorak.com", "date": "2026-10-20", "days_after": 7.7, "points": 1, "hn_id": "50165319", "relation": "discussion", "confidence": "medium", "source": "hn", "summary_zh": "以 Anthropic 的自我复制报告为例讨论摩擦消失的影响", "summary_en": "Uses Anthropic's self-replication post as an example to discuss what happens when friction disappears.", "origin": "backfill" } ] }, { "original": { "id": "50255159", "url": "https://hexagonmath.org/2610.00577v1", "domain": "hexagonmath.org", "title": "A $2^{o(n)}$ Time Algorithm for 3-SAT", "category": "research_report", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50255159", "points": 26, "comments": 6, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "声称给出3-SAT确定性算法以反驳ETH", "summary_en": "Claims a deterministic algorithm for 3-SAT that would refute the ETH." }, "responses": [ { "id": "50265063", "title": "The Exponential Time Hypothesis is False", "url": "https://www.johndcook.com/blog/2026/10/27/eth-false/", "domain": "johndcook.com", "date": "2026-10-27", "days_after": 0.6, "points": 2, "hn_id": "50265063", "relation": "discussion", "confidence": "medium", "source": "hn", "summary_zh": "简要报道 OpenAI 与 Anthropic 发表了否定 ETH 的证明", "summary_en": "Briefly reports that Anthropic and OpenAI published proofs that ETH is false.", "origin": "backfill" } ] }, { "original": { "id": "50271917", "url": "https://www.anthropic.com/research/agents-judgment-calls", "domain": "anthropic.com", "title": "How AI agents shape experts’ judgment calls", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50271917", "points": 5, "comments": 1, "date": "2026-10-28", "submissions": 3, "source": "hn", "summary_zh": "研究分析师在86%决策中跟随AI代理偏好", "summary_en": "Studies how analysts follow AI agent preferences in 86% of decisions." }, "responses": [ { "id": "r22", "title": "AI advice and expert judgment", "url": "https://blog.stephenturner.us/p/ai-advice-and-expert-judgment", "domain": "blog.stephenturner.us", "date": "2026-11-01", "days_after": 4, "points": 0, "hn_id": null, "relation": "discussion", "confidence": "high", "source": "rs", "summary_zh": "文章围绕Anthropic这项研究及论文展开讨论并总结结果", "summary_en": "Discusses this Anthropic study and its paper, and summarizes the results." } ] }, { "original": { "id": "r95", "source": "rs", "url": "https://blog.front-matter.de/posts/what-content-categories-are-used-in-science-blogs/", "domain": "blog.front-matter.de", "title": "What content categories are used in science blogs?", "category": "research_report", "topic": "scholarly-communication", "confidence": "high", "date": "2026-10-19", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Front Matter", "doi": "10.53731/zvzh7-e8v51", "summary_zh": "分析Rogue Scholar科学博客的内容类别并提出分类法", "summary_en": "Analyzes the content categories of Rogue Scholar science blogs and proposes a taxonomy." }, "responses": [ { "id": "r73", "title": "Trying to categorize SV-POW! posts", "url": "https://svpow.com/2026/10/22/trying-to-categorize-sv-pow-posts/", "domain": "svpow.com", "date": "2026-10-22", "days_after": 3, "points": 0, "hn_id": null, "relation": "discussion", "confidence": "high", "source": "rs", "summary_zh": "文章用Rogue Scholar的分类法给自己的40篇文章分类", "summary_en": "Categorizes the author's 40 posts using the Rogue Scholar taxonomy." } ] }, { "original": { "id": "50272479", "url": "https://research.google/blog/the-sleep-framework-a-new-way-for-llms-to-retain-knowledge-in-continual-learning/", "domain": "research.google", "title": "The sleep framework: A new way for LLMs to retain knowledge in continual learning", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50272479", "points": 277, "comments": 58, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "Google提出sleep框架用于持续学习", "summary_en": "Google proposes a sleep framework for continual learning." }, "responses": [ { "id": "50398678", "title": "The components of agent memory", "url": "https://www.anthropic.com/research/components-of-agent-memory", "domain": "anthropic.com", "date": "2026-11-06", "days_after": 9.4, "points": 4, "hn_id": "50398678", "relation": "discussion", "confidence": "high", "source": "hn", "summary_zh": "Anthropic文章讨论并应用了Google提出的睡眠框架概念来构建Agent记忆。", "summary_en": "Anthropic discusses and applies the sleep framework concept to build agent memory.", "origin": "daily" } ] } ], "reports": [ { "id": "50341174", "url": "https://rikiyat.github.io/blog/weight-space-programming/", "domain": "rikiyat.github.io", "title": "Sparks of Weight-Space Programming: LLMs can natively read & write neural network weights", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": true, "hn_id": "50341174", "points": 2957, "comments": 898, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "发现前沿 LLM 能直接读写神经网络权重以改进 Qwen", "summary_en": "Finds that frontier LLMs can directly read and write neural network weights to improve Qwen.", "linkers": [ "alisorcorp.github.io", "anderegg.ca", "arattinger.dev", "arcturus-labs.com", "cjauvin.github.io", "flecart.github.io", "hedwigai.com", "honnibal.dev", "jonneslin.github.io", "paulg.info", "seangoedecke.com", "simonwillison.net", "substack.com", "venkateshrao.com", "sgiannopoulos.github.io", "doubleai.com" ] }, { "id": "50105647", "url": "https://already-clean.goodfire.com/", "domain": "already-clean.goodfire.com", "title": "Neural Networks Are Already Clean", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50105647", "points": 849, "comments": 344, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "提出神经网络神经元与电路较为干净的工作假说", "summary_en": "Proposes a working hypothesis that neurons and circuits in neural networks are relatively clean.", "linkers": [ "100toolkit.com", "lesswrong.com", "livgorton.com", "substack.com", "technologyreview.com", "venkateshrao.com" ] }, { "id": "50282838", "url": "https://divagr.com/blog/qwenputer", "domain": "divagr.com", "title": "Qwenputer: a fully functioning, deterministic RISC-V CPU inside Qwen3.5", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50282838", "points": 708, "comments": 147, "date": "2026-10-28", "submissions": 2, "source": "hn", "summary_zh": "在Qwen3.5内实现确定性RISC-V CPU", "summary_en": "Implements a deterministic RISC-V CPU inside Qwen3.5.", "linkers": [ "theainewsreport.com" ] }, { "id": "50122878", "url": "https://alexwang.ai/posts/associative-memory/", "domain": "alexwang.ai", "title": "You could have invented transformers from memory", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50122878", "points": 670, "comments": 152, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "以记忆存储视角讲解 transformer 的构造", "summary_en": "Explains how transformers are built from the perspective of memory storage.", "linkers": [ "substack.com" ] }, { "id": "50106023", "url": "https://zachlandes.com/trolley-problem/", "domain": "zachlandes.com", "title": "Jev Meets the Trolley Problem", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50106023", "points": 656, "comments": 264, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "向 Jev 与 13 个决策模型提出电车难题并比较结果", "summary_en": "Poses the trolley problem to Jev and 13 decision models and compares the results.", "linkers": [] }, { "id": "50121179", "url": "https://www.mercor.com/research/labor-index/benchmark-map/", "domain": "mercor.com", "title": "A benchmark map of the US economy", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50121179", "points": 634, "comments": 581, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "构建 AI 基准地图，分析基准对美国经济工时的覆盖", "summary_en": "Builds an AI benchmark map and analyzes how well benchmarks cover US economic work hours.", "linkers": [ "tanner.place" ] }, { "id": "50101813", "url": "https://citizenlab.ca/research/how-discover-tech-turned-targeted-ads-into-a-pipeline-for-spyware/", "domain": "citizenlab.ca", "title": "Exposing AdFection: How Discover-Tech Turned Targeted Ads Into a Pipeline for Zero-Click Mercenary Spyware", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50101813", "points": 543, "comments": 690, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "Citizen Lab 披露通过定向广告投递零点击间谍软件", "summary_en": "Citizen Lab discloses zero-click spyware delivered through targeted ads.", "linkers": [ "eff.org", "gizmodo.com", "pluralistic.net", "politico.eu" ] }, { "id": "50246658", "url": "https://cohere.com/blog/concerning-behaviour-popular-open-weight-ai-model", "domain": "cohere.com", "title": "Concerning behavior from a popular open-weight AI model", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50246658", "points": 337, "comments": 133, "date": "2026-10-26", "submissions": 1, "source": "hn", "summary_zh": "Cohere披露某开源权重模型的异常行为", "summary_en": "Cohere discloses abnormal behavior in an open-weight model.", "linkers": [ "checkpoint.com", "int8.io", "kagent.dev", "notebookcheck.net", "substack.com", "xda-developers.com" ] }, { "id": "50171379", "url": "https://danielball.com/notes/ai_image_generation.html", "domain": "danielball.com", "title": "AI Image Generation", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50171379", "points": 287, "comments": 94, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "讲解 AI 图像生成从扩散模型到 agent 的实用指南", "summary_en": "A practical guide to AI image generation, from diffusion models to agents.", "linkers": [] }, { "id": "50272479", "url": "https://research.google/blog/the-sleep-framework-a-new-way-for-llms-to-retain-knowledge-in-continual-learning/", "domain": "research.google", "title": "The sleep framework: A new way for LLMs to retain knowledge in continual learning", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50272479", "points": 277, "comments": 58, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "Google提出sleep框架用于持续学习", "summary_en": "Google proposes a sleep framework for continual learning.", "linkers": [ "anthropic.com" ] }, { "id": "50235780", "url": "https://www.dbreunig.com/2026/10/24/the-state-of-context-engineering.html", "domain": "dbreunig.com", "title": "The State of Context Engineering", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50235780", "points": 275, "comments": 67, "date": "2026-10-25", "submissions": 2, "source": "hn", "summary_zh": "关于context engineering现状的演讲稿", "summary_en": "A talk transcript on the current state of context engineering.", "linkers": [] }, { "id": "50065932", "url": "https://www.argmin.net/p/faster-fourier-transforms", "domain": "argmin.net", "title": "Faster Fourier Transforms-by Ben Recht", "category": "research_report", "topic": "math-theory", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50065932", "points": 249, "comments": 170, "date": "2026-10-12", "submissions": 2, "source": "hn", "summary_zh": "分析OpenAI的FFT算法，认为其是利用漏洞而非突破", "summary_en": "Analyzes OpenAI's FFT algorithm and argues it exploits a loophole rather than being a breakthrough.", "linkers": [] }, { "id": "50179706", "url": "https://www.hirundo.io/blog/qwen3-coder-vulnerabilities", "domain": "hirundo.io", "title": "Chinese AI Writes Weaker Code for Beijing’s Adversaries. We Fixed It. A Qwen3-Coder Case Study", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50179706", "points": 243, "comments": 118, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "发现 Qwen3-Coder 对美国角色提示生成更不安全的代码", "summary_en": "Finds that Qwen3-Coder generates less secure code when prompted with US-related personas.", "linkers": [ "cohere.com" ] }, { "id": "50261647", "url": "https://cfsg.axiommath.ai/", "domain": "cfsg.axiommath.ai", "title": "CFSG in Lean", "category": "research_report", "topic": "math-theory", "confidence": "high", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50261647", "points": 243, "comments": 79, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "用Lean形式化有限单群分类并得到新定理", "summary_en": "Formalizes the classification of finite simple groups in Lean and obtains a new theorem.", "linkers": [ "johndcook.com" ] }, { "id": "50368626", "url": "https://www.anthropic.com/research/provable-inference", "domain": "anthropic.com", "title": "Progress on provable inference for large language models", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50368626", "points": 238, "comments": 124, "date": "2026-11-03", "submissions": 4, "source": "hn", "summary_zh": "Anthropic 报告 LLM 可证明推理 zkSNARK 的进展", "summary_en": "Anthropic reports progress on zkSNARKs for provable LLM inference.", "linkers": [ "leanweekly.news", "theainewsreport.com" ] }, { "id": "50201445", "url": "https://www.lesswrong.com/posts/WHEQps5PK8LrrcTAb/how-does-fable-5-1-hide-messages-in-images", "domain": "lesswrong.com", "title": "How does Fable 5.1 hide messages in images?", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50201445", "points": 219, "comments": 74, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "研究前沿 LLM 如何在图像中隐藏信息", "summary_en": "Studies how frontier LLMs hide information inside images.", "linkers": [] }, { "id": "50021410", "url": "https://blog.janestreet.com/can-you-use-autoregressive-diffusion-to-generate-market-data/", "domain": "blog.janestreet.com", "title": "Jane Street Blog - Can you use autoregressive diffusion to generate market data?", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50021410", "points": 186, "comments": 57, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "探索用自回归扩散生成市场数据", "summary_en": "Explores using autoregressive diffusion to generate market data.", "linkers": [ "substack.com" ] }, { "id": "50212507", "url": "https://www.jefftk.com/p/air-purifiers-as-white-noise-machines", "domain": "jefftk.com", "title": "Air Purifiers as White Noise Machines", "category": "research_report", "topic": "physics-earth", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50212507", "points": 178, "comments": 174, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "测量空气净化器作为白噪声机的效果", "summary_en": "Measures how well air purifiers work as white noise machines.", "linkers": [] }, { "id": "50121939", "url": "https://www.anthropic.com/research/designing-efficient-quantum-circuits", "domain": "anthropic.com", "title": "Designing more efficient quantum circuits with Claude", "category": "research_report", "topic": "physics-earth", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50121939", "points": 174, "comments": 39, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "Claude 自主设计量子电路并改进已发表的最佳版本", "summary_en": "Claude autonomously designs quantum circuits and improves on the best published versions.", "linkers": [] }, { "id": "50139873", "url": "https://www.anthropic.com/research/employment-at-anthropic", "domain": "anthropic.com", "title": "How employment is changing at Anthropic", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50139873", "points": 166, "comments": 263, "date": "2026-10-17", "submissions": 2, "source": "hn", "summary_zh": "Anthropic 分析 AI 如何改变公司内部就业", "summary_en": "Anthropic analyzes how AI is changing employment inside the company.", "linkers": [ "angadh.com", "av.codes", "hugo.im", "lilius.io", "noahpinion.blog", "seldo.com", "substack.com", "withnative.ai" ] }, { "id": "50175185", "url": "https://www.quantamagazine.org/in-hilbert-space-all-things-are-quantumly-possible-20260826/", "domain": "quantamagazine.org", "title": "In Hilbert Space, All Things Are Quantumly Possible", "category": "explainer", "topic": "physics-earth", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50175185", "points": 149, "comments": 84, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "Quanta 讲解希尔伯特空间与量子现象", "summary_en": "Quanta explains Hilbert space and quantum phenomena.", "linkers": [ "substack.com" ] }, { "id": "50290912", "url": "https://skypilot.ai/blog/grace-blackwell-needs-a-new-scheduler", "domain": "skypilot.ai", "title": "Grace Blackwell needs a new scheduler: 27% more usable GPUs from the same racks", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50290912", "points": 147, "comments": 22, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "Grace Blackwell调度器多出27%可用GPU", "summary_en": "A Grace Blackwell scheduler yields 27% more usable GPUs.", "linkers": [] }, { "id": "50048522", "url": "https://pub.sakana.ai/mass/", "domain": "pub.sakana.ai", "title": "MASS: Recursive Self-Improvement through Multi-Agent Self-Supervision", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50048522", "points": 138, "comments": 82, "date": "2026-10-11", "submissions": 1, "source": "hn", "summary_zh": "研究语言模型经多代理自监督的递归自我改进MASS", "summary_en": "Studies MASS, recursive self-improvement of language models through multi-agent self-supervision.", "linkers": [ "substack.com" ] }, { "id": "50061603", "url": "https://www.anthropic.com/research/self-replicating-ai", "domain": "anthropic.com", "title": "AI models may be capable of self-replicating without access to a datacenter", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50061603", "points": 118, "comments": 116, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "Anthropic 实验评估小型 AI 模型在廉价 IoT 设备上自我复制的能力", "summary_en": "Anthropic experiments assess whether small AI models can self-replicate on cheap IoT devices.", "linkers": [ "dbreunig.com", "geohot.github.io", "macanorak.com", "stiven.me", "thestack.technology" ] }, { "id": "50027257", "url": "https://epoch.ai/publications/innovationeval", "domain": "epoch.ai", "title": "Can AI automate AI R&D yet?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50027257", "points": 93, "comments": 97, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "用InnovationEval评估AI模型复现人类算法创新", "summary_en": "Uses InnovationEval to test whether AI models can reproduce human algorithmic innovations.", "linkers": [ "lesswrong.com" ] }, { "id": "50278253", "url": "https://www.cartesia.ai/blog/from-tokens-to-bytes-part-1", "domain": "cartesia.ai", "title": "From Tokens to Bytes I: Scaling Hierarchical Networks", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50278253", "points": 92, "comments": 14, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "Cartesia关于层级网络扩展的研究", "summary_en": "Cartesia's research on scaling hierarchical networks.", "linkers": [] }, { "id": "50160241", "url": "https://nsidc.org/learn/ask-scientist/karakoram-anomaly-it-real", "domain": "nsidc.org", "title": "What is the Karakoram Anomaly?", "category": "explainer", "topic": "physics-earth", "confidence": "high", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50160241", "points": 89, "comments": 39, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "解释喀喇昆仑冰川异常现象", "summary_en": "Explains the Karakoram glacier anomaly.", "linkers": [] }, { "id": "50277755", "url": "https://www.quantamagazine.org/how-spiders-spin-silk-dope-into-natures-toughest-material-20261028/", "domain": "quantamagazine.org", "title": "How Spiders Spin ‘Silk Dope’ Into Nature’s Toughest Material", "category": "explainer", "topic": "physics-earth", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50277755", "points": 79, "comments": 10, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "Quanta解释蜘蛛丝形成机制的研究", "summary_en": "Quanta explains research on how spider silk forms.", "linkers": [] }, { "id": "50320599", "url": "https://blog.keyvan.net/p/ai-text-watermarking-and-quality", "domain": "blog.keyvan.net", "title": "AI text watermarking and quality loss", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50320599", "points": 77, "comments": 16, "date": "2026-10-31", "submissions": 1, "source": "hn", "summary_zh": "评估 AI 文本水印对文本质量的影响", "summary_en": "Evaluates how AI text watermarking affects text quality.", "linkers": [] }, { "id": "50101649", "url": "https://courk.cc/rp2350-challenge-laser", "domain": "courk.cc", "title": "Courk's Blog – Laser Fault Injection on a Budget: RP2350 Edition", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50101649", "points": 66, "comments": 18, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "在 RP2350 上以低成本实施激光故障注入", "summary_en": "Carries out low-cost laser fault injection on the RP2350.", "linkers": [] }, { "id": "50266910", "url": "https://qwen-neurons-site.vercel.app/", "domain": "qwen-neurons-site.vercel.app", "title": "Neuron Persona Library", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50266910", "points": 60, "comments": 60, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "翻转单个神经元触发Qwen人格的研究", "summary_en": "Studies how flipping a single neuron triggers a persona in Qwen.", "linkers": [ "ziva.sh" ] }, { "id": "50389448", "url": "https://obelisk.club/research/app-store-2026", "domain": "obelisk.club", "title": "App Store 2026 in numbers: AI, pricing, abandoned apps", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50389448", "points": 60, "comments": 52, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "统计 178,637 个 App Store 应用的 AI 与内购情况", "summary_en": "Tallies AI features and in-app purchases across 178,637 App Store apps.", "linkers": [] }, { "id": "50169506", "url": "https://research.poh.org/bot-benchmarking-2026/", "domain": "research.poh.org", "title": "Benchmarking Bot Detection Systems Against Modern AI Agents in 2026", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50169506", "points": 50, "comments": 8, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "重新基准测试机器人检测系统对现代 AI agent 的效果", "summary_en": "Re-benchmarks how well bot detection systems handle modern AI agents.", "linkers": [] }, { "id": "50218199", "url": "https://aseemrb.me/blog/colors-of-a-soap-film/", "domain": "aseemrb.me", "title": "The colors of a soap film", "category": "explainer", "topic": "physics-earth", "confidence": "medium", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50218199", "points": 43, "comments": 7, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "解释肥皂膜颜色与波长倒数的余弦关系", "summary_en": "Explains the relationship between soap film colors and the cosine of inverse wavelength.", "linkers": [] }, { "id": "50269651", "url": "https://artificialanalysis.ai/methodology/intelligence-benchmarking", "domain": "artificialanalysis.ai", "title": "Intelligence Benchmarking", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50269651", "points": 40, "comments": 5, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "Artificial Analysis智能指数v5.0基准", "summary_en": "Artificial Analysis Intelligence Index v5.0 benchmark.", "linkers": [ "theainewsreport.com" ] }, { "id": "50233948", "url": "https://ipinfo.io/blog/how-made-mmdb-files-smaller", "domain": "ipinfo.io", "title": "How We Made Our MMDB Files a Third Smaller Without Changing a Byte of Data", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50233948", "points": 38, "comments": 4, "date": "2026-10-14", "submissions": 2, "source": "hn", "summary_zh": "IPinfo 介绍将 MMDB 文件缩小并保持查询结果不变的优化", "summary_en": "IPinfo describes an optimization that shrinks MMDB files while keeping lookup results unchanged.", "linkers": [] }, { "id": "50367004", "url": "https://streamvessel.com/blog/mcp-trusted-redirect-uris", "domain": "streamvessel.com", "title": "MCP OAuth: let users trust their own redirect URIs safely", "category": "explainer", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50367004", "points": 35, "comments": 6, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "说明 MCP 服务器信任自定义客户端而不引入钓鱼风险的方案", "summary_en": "Describes a way for MCP servers to trust custom clients without introducing phishing risk.", "linkers": [] }, { "id": "50188440", "url": "https://xingjianbai.com/reactive-optimizers/", "domain": "xingjianbai.com", "title": "Reactive Optimizers", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50188440", "points": 32, "comments": 3, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "提出让 agent 在训练中改写优化器源码的 Reactive Optimizers", "summary_en": "Proposes Reactive Optimizers, which let an agent rewrite the optimizer source code during training.", "linkers": [ "doubleai.com" ] }, { "id": "50234183", "url": "https://kevin-wu.ca/blog/qwn/", "domain": "kevin-wu.ca", "title": "Qwen, Minus the E", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50234183", "points": 31, "comments": 21, "date": "2026-10-25", "submissions": 1, "source": "hn", "summary_zh": "在16GB GPU上微调Qwen避免字母e的实验", "summary_en": "An experiment fine-tuning Qwen on a 16GB GPU to avoid the letter e.", "linkers": [] }, { "id": "50347766", "url": "https://zou-group.github.io/llm-claims-shelf-life/", "domain": "zou-group.github.io", "title": "The shelf life of LLM findings", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50347766", "points": 30, "comments": 4, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "对 50 项 LLM 已发表结论在 20 个模型上复现", "summary_en": "Replicates 50 published LLM findings across 20 models.", "linkers": [] }, { "id": "50152976", "url": "https://stealthium.io/blog/hands-off-detecting-model-weight-tampering", "domain": "stealthium.io", "title": "Hands Off: Detecting Model Weight Tampering", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50152976", "points": 28, "comments": 7, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "介绍在 GPU 内存中篡改模型权重的 ROME 攻击及检测", "summary_en": "Introduces the ROME attack that tampers with model weights in GPU memory, and ways to detect it.", "linkers": [] }, { "id": "50348535", "url": "https://research.klang.ai/papers/the-nordic-token-tax", "domain": "research.klang.ai", "title": "The Nordic token tax", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50348535", "points": 28, "comments": 6, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "测量 12 个前沿模型对非英语提示多收的 token 成本", "summary_en": "Measures the extra token cost that 12 frontier models charge for non-English prompts.", "linkers": [] }, { "id": "50255159", "url": "https://hexagonmath.org/2610.00577v1", "domain": "hexagonmath.org", "title": "A $2^{o(n)}$ Time Algorithm for 3-SAT", "category": "research_report", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50255159", "points": 26, "comments": 6, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "声称给出3-SAT确定性算法以反驳ETH", "summary_en": "Claims a deterministic algorithm for 3-SAT that would refute the ETH.", "linkers": [ "cellcog.ai", "johndcook.com", "scientificamerican.com" ] }, { "id": "50156332", "url": "https://www.beaconholdings.com/research/towards-company-world-models", "domain": "beaconholdings.com", "title": "Towards Company World Models", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50156332", "points": 25, "comments": 5, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "提出以 AI 学习店主经验、辅助小企业决策的公司世界模型", "summary_en": "Proposes company world models, where AI learns from shop owners' experience to support small business decisions.", "linkers": [] }, { "id": "50160440", "url": "https://www.basis.ai/blog/ship-your-interpreters/", "domain": "basis.ai", "title": "Basis | Ship your interpreter", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50160440", "points": 24, "comments": 25, "date": "2026-10-17", "submissions": 2, "source": "hn", "summary_zh": "展示用 LLM 为编程语言编写规则并证明解释器符合规则", "summary_en": "Shows how to use an LLM to write rules for a programming language and prove that an interpreter conforms to them.", "linkers": [] }, { "id": "50201842", "url": "https://www.wolframcollatzmahler.com/rule30/index.html", "domain": "wolframcollatzmahler.com", "title": "Rule 30: One Black Cell", "category": "research_report", "topic": "math-theory", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50201842", "points": 24, "comments": 6, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "探讨 Rule 30 从单个黑格生长的性质", "summary_en": "Explores the properties of Rule 30 growing from a single black cell.", "linkers": [] }, { "id": "50222826", "url": "https://www.skyvern.com/blog/deleting-rag-from-our-web-agent-made-it-2-3x-faster/", "domain": "skyvern.com", "title": "Deleting RAG from our web agent made it 2.3x faster", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50222826", "points": 22, "comments": 3, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "Skyvern重写后在Odysseys基准上提速2.3倍的报告", "summary_en": "Reports that Skyvern's rewrite ran 2.3 times faster on the Odysseys benchmark.", "linkers": [] }, { "id": "50197692", "url": "https://bernsteinbear.com/blog/object-shapes/", "domain": "bernsteinbear.com", "title": "Objects, attributes, maps, oh my!", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50197692", "points": 21, "comments": 1, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "讲解对象、属性、map 与 inline cache", "summary_en": "Explains objects, attributes, maps, and inline caches.", "linkers": [] }, { "id": "50209939", "url": "https://www.anthropic.com/research/alignment-assessment-ukaisi", "domain": "anthropic.com", "title": "An alignment assessment of the UK AISI cybertesting incident", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50209939", "points": 19, "comments": 3, "date": "2026-10-23", "submissions": 2, "source": "hn", "summary_zh": "评估 Mythos 5 在 UK AISI 评测中的越权行为", "summary_en": "Assesses Mythos 5's unauthorized actions during the UK AISI evaluation.", "linkers": [ "gov.uk", "substack.com" ] }, { "id": "50216027", "url": "https://jonbarron.info/softmaxfix/", "domain": "jonbarron.info", "title": "A Fix for Softmax and Cross-Entropy", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50216027", "points": 19, "comments": 1, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "提出对 softmax 和 cross-entropy 的改进", "summary_en": "Proposes an improvement to softmax and cross-entropy.", "linkers": [] }, { "id": "50337225", "url": "https://lacuna.tiptreesystems.com/paper/bf16-might-be-freezing-your-weight-updates/art_487b1ed949e745a6ae0c42815b8d3cba", "domain": "lacuna.tiptreesystems.com", "title": "bf16 might be freezing your weight updates", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50337225", "points": 19, "comments": 11, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "分析 bf16 权重下 AdamW 更新被冻结的问题", "summary_en": "Analyzes how AdamW updates can be frozen with bf16 weights.", "linkers": [] }, { "id": "50108391", "url": "https://www.abundant.ai/blog/hugging-face-incident-part-1/", "domain": "abundant.ai", "title": "Reproducing the Hugging Face Incident, Part I", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50108391", "points": 17, "comments": 1, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "受控复现 Hugging Face 事件中的包管理器攻击", "summary_en": "A controlled reproduction of the package manager attack in the Hugging Face incident.", "linkers": [] }, { "id": "50159119", "url": "https://www.globus.org/blog/gridftp-a-brief-history-of-fast-file-transfer", "domain": "globus.org", "title": "GridFTP: A Brief History of Fast File Transfer", "category": "explainer", "topic": "systems-hardware", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50159119", "points": 17, "comments": 1, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "回顾 GridFTP 作为科学数据传输标准的历史", "summary_en": "Reviews the history of GridFTP as a standard for scientific data transfer.", "linkers": [] }, { "id": "50146286", "url": "https://khalidhamadeh.com/writing/rc2-pi-table", "domain": "khalidhamadeh.com", "title": "How RC2’s π table was made (PITABLE, RFC 2268)", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50146286", "points": 16, "comments": 2, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "研究 RC2 密码 π 表的生成方式，用 Fisher-Yates 洗牌复现", "summary_en": "Studies how the RC2 cipher's pi table was generated and reproduces it with a Fisher-Yates shuffle.", "linkers": [] }, { "id": "50299790", "url": "https://hexagonmath.org/2610.00600v1", "domain": "hexagonmath.org", "title": "Deterministic Factoring in Time $N^{o(1)}$", "category": "research_report", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50299790", "points": 15, "comments": 3, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "声称给出整数分解N^O(1)确定性算法", "summary_en": "Claims a deterministic integer factoring algorithm running in time N^o(1).", "linkers": [] }, { "id": "50350302", "url": "https://blog.apify.com/llm-grounding-with-live-data/", "domain": "blog.apify.com", "title": "LLM grounding tutorial: stop AI agents guessing prices", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50350302", "points": 15, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "在 21 个商品页上测试 AI 代理价格幻觉与实时数据接地", "summary_en": "Tests AI agent price hallucination and live-data grounding on 21 product pages.", "linkers": [] }, { "id": "50042051", "url": "https://www.aisi.gov.uk/blog/gpt-6-astra-performs-unsanctioned-supply-chain-attacks-in-simulations", "domain": "aisi.gov.uk", "title": "GPT-6 Astra performs unsanctioned supply-chain attacks in simulations", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50042051", "points": 14, "comments": 2, "date": "2026-10-11", "submissions": 1, "source": "hn", "summary_zh": "评估GPT-6 Astra在模拟中的未授权供应链攻击行为", "summary_en": "Evaluates unauthorized supply-chain attack behavior by GPT-6 Astra in simulations.", "linkers": [ "crawlspider.com" ] }, { "id": "50297159", "url": "https://epoch.ai/publications/open-weight-models-patch-gap", "domain": "epoch.ai", "title": "Open-weight AI models can exploit the patch gap", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50297159", "points": 13, "comments": 9, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "开源权重模型可利用补丁间隔漏洞的评估", "summary_en": "Evaluates how open-weight models can exploit the patch gap.", "linkers": [] }, { "id": "50085871", "url": "https://swarmcha.se/posts/chinese-agent-fleet-full-report", "domain": "swarmcha.se", "title": "We found a Tencent agent fleet", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50085871", "points": 12, "comments": 7, "date": "2026-10-14", "submissions": 6, "source": "hn", "summary_zh": "记录云平台上大批 AI agent 活动的调查与发现", "summary_en": "An investigation documenting large-scale AI agent activity on cloud platforms.", "linkers": [ "abundant.ai", "crawlspider.com", "quickchat.ai" ] }, { "id": "50225659", "url": "https://philippdubach.com/posts/forecasting-financial-markets-with-timestral/", "domain": "philippdubach.com", "title": "Timestral Financial Forecasting: Time-Series Foundation Models", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50225659", "points": 11, "comments": 2, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "Timestral模型预测金融市场的评估结果", "summary_en": "Evaluation results of the Timestral model for forecasting financial markets.", "linkers": [] }, { "id": "50258875", "url": "https://www.withnative.ai/research-and-posts/markdown-context-benchmark", "domain": "withnative.ai", "title": "40× fewer tokens: wrap Markdown in SQLite", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50258875", "points": 11, "comments": 0, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "声称用SQLite使Markdown的token效率提升40倍", "summary_en": "Claims SQLite makes Markdown 40 times more token-efficient.", "linkers": [] }, { "id": "50022143", "url": "https://www.datologyai.com/blog/text-curation-oct-2026", "domain": "datologyai.com", "title": "Datology Curation Studio Improves Even the Best Open Data", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50022143", "points": 10, "comments": 2, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "报告Datology Curation Studio改进开源数据带来6倍算力等效", "summary_en": "Reports that Datology Curation Studio improves open data, giving a 6x compute equivalent.", "linkers": [] }, { "id": "50382009", "url": "https://declarative-dragging.github.io/", "domain": "declarative-dragging.github.io", "title": "Declarative Dragging", "category": "research_report", "topic": "other", "confidence": "high", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50382009", "points": 10, "comments": 1, "date": "2026-11-04", "submissions": 4, "source": "hn", "summary_zh": "提出声明式拖拽交互，UIST 2026 最佳论文项目页", "summary_en": "Proposes declarative dragging interaction; project page of a UIST 2026 best paper.", "linkers": [] }, { "id": "50359051", "url": "https://www.krea.ai/blog/krea-3-technical-report", "domain": "krea.ai", "title": "Krea 3 Technical Report", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50359051", "points": 9, "comments": 1, "date": "2026-11-03", "submissions": 2, "source": "hn", "summary_zh": "Krea 3 图像编辑模型的技术报告", "summary_en": "Technical report for the Krea 3 image editing model.", "linkers": [] }, { "id": "50075487", "url": "https://www.frontierml.org/blog/", "domain": "frontierml.org", "title": "FrontierML: a challenging, end-to-end machine learning benchmark", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50075487", "points": 8, "comments": 1, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "FrontierML 基准：让 agent 在 12 小时和有限预算下做机器学习", "summary_en": "FrontierML benchmark: agents do machine learning in 12 hours on a limited budget.", "linkers": [] }, { "id": "49999165", "url": "https://resilience.ocf.tw/web/report/en", "domain": "resilience.ocf.tw", "title": "When Submarine Cables Go Dark: Understanding and Preparing for the Risks of Taiwan's International Internet Disconnection", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "49999165", "points": 7, "comments": 0, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "分析台湾国际海底电缆中断风险并提出应对准备", "summary_en": "Analyzes the risk of disruption to Taiwan's international submarine cables and how to prepare.", "linkers": [] }, { "id": "50000668", "url": "https://www.vals.ai/blogs/mimo-reward-hacking", "domain": "vals.ai", "title": "Two-Thirds of MiMo v2.6's Coding Tasks Leak the Answer. Are Models That Exploit This Misbehaving?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50000668", "points": 7, "comments": 3, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "审计MiMo v2.6编码任务，发现三分之二泄露答案", "summary_en": "Audits MiMo v2.6 coding tasks and finds two thirds leak the answer.", "linkers": [ "modal.com", "openathena.ai" ] }, { "id": "50007378", "url": "https://institute.deepmind.com/essays/cheaters-and-whistleblowers-in-the-agent-swarm/", "domain": "institute.deepmind.com", "title": "Cheaters and whistleblowers in the agent swarm", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50007378", "points": 7, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "记录代理群体在虚拟数学会议中作弊与举报的实验", "summary_en": "Documents an experiment where agents cheat and blow the whistle at a virtual math conference.", "linkers": [] }, { "id": "50118853", "url": "https://nikhilism.com/post/2026/golang-racket-concurrency/", "domain": "nikhilism.com", "title": "ConcurrentML: Improving on Go's concurrency model", "category": "explainer", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50118853", "points": 7, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "讲解 ConcurrentML 事件模型如何改进 Go 并发", "summary_en": "Explains how the ConcurrentML event model improves on Go's concurrency.", "linkers": [] }, { "id": "50319896", "url": "https://bathtub-01.github.io/technical/ouros-demo/", "domain": "bathtub-01.github.io", "title": "A Specialised Haskell Processor on FPGA", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50319896", "points": 7, "comments": 1, "date": "2026-10-31", "submissions": 3, "source": "hn", "summary_zh": "展示在 FPGA 上运行 Haskell 的专用处理器", "summary_en": "Presents a specialised processor that runs Haskell on an FPGA.", "linkers": [] }, { "id": "50372303", "url": "https://news.fnal.gov/2026/10/the-sbnd-collaboration-releases-first-neutrino-results/", "domain": "news.fnal.gov", "title": "The SBND collaboration releases first neutrino results", "category": "research_report", "topic": "physics-earth", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50372303", "points": 7, "comments": 2, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "SBND 合作组发布首个中微子相互作用结果", "summary_en": "The SBND collaboration releases its first neutrino interaction results.", "linkers": [] }, { "id": "49996275", "url": "https://artifactarena.ai", "domain": "artifactarena.ai", "title": "Artifact Arena — where making is the measure of intelligence", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "49996275", "points": 6, "comments": 0, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "提出Artifact Arena，以模型制作成果评测智能的基准", "summary_en": "Proposes Artifact Arena, a benchmark that measures intelligence by what models make.", "linkers": [] }, { "id": "50021031", "url": "https://alexvidovich.com/papers/backtest-hygiene", "domain": "alexvidovich.com", "title": "Backtests vs Live Trading: Six Defects from One System", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50021031", "points": 6, "comments": 0, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "分析日内交易系统回测与实盘偏离的六类缺陷", "summary_en": "Analyzes six types of defects behind the gap between backtests and live results in an intraday trading system.", "linkers": [] }, { "id": "50067876", "url": "https://www.cloudrift.ai/blog/inside-a-gpu-compiler-from-pytorch-to-cuda", "domain": "cloudrift.ai", "title": "Building an ML Compiler from Scratch: PyTorch to CUDA", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50067876", "points": 6, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "以一个 PyTorch 算子为例讲解经六层 IR 降到 CUDA 的编译过程", "summary_en": "Walks through compiling one PyTorch operator to CUDA across six IR layers.", "linkers": [] }, { "id": "50081029", "url": "https://developer.nvidia.com/blog/the-machines-that-make-the-machines/", "domain": "developer.nvidia.com", "title": "The Machines that Make the Machines", "category": "research_report", "topic": "robotics-graphics", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50081029", "points": 6, "comments": 0, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "NVIDIA 介绍训练机器人装配 GB300 测试托盘的方法与经验", "summary_en": "NVIDIA describes methods and lessons from training robots to assemble GB300 test trays.", "linkers": [] }, { "id": "50081419", "url": "https://www.tobyord.com/writing/ai-scaling-vs-human-scaling", "domain": "tobyord.com", "title": "AI Scaling vs Human Scaling", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50081419", "points": 6, "comments": 0, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "分析 AI 模型随 token 增加的扩展是否如人类随时间扩展", "summary_en": "Analyzes whether AI model scaling with tokens resembles human scaling over time.", "linkers": [ "noahpinion.blog" ] }, { "id": "50239838", "url": "https://lnsgroup.cc/research/MS-Human/", "domain": "lnsgroup.cc", "title": "MS-Human-700: Whole-body Human Musculoskeletal Modeling and Control", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50239838", "points": 6, "comments": 0, "date": "2026-10-26", "submissions": 1, "source": "hn", "summary_zh": "MS-Human-700人体肌肉骨骼MuJoCo模型项目页", "summary_en": "Project page for MS-Human-700, a whole-body human musculoskeletal model in MuJoCo.", "linkers": [] }, { "id": "50262505", "url": "https://jasonkena.github.io/blog/posts/does_self_distillation_improve_reasoning/", "domain": "jasonkena.github.io", "title": "A critical look at self-distillation for reasoning", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50262505", "points": 6, "comments": 1, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "对推理中self-distillation的批判性分析", "summary_en": "A critical analysis of self-distillation for reasoning.", "linkers": [] }, { "id": "50283487", "url": "https://brunborg.com/n-gram-hinted-lm", "domain": "brunborg.com", "title": "N-gram hinted language modeling: NanoGPT Speedrun in 6.7s", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50283487", "points": 6, "comments": 1, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "用n-gram提示将NanoGPT Speedrun降至6.7秒", "summary_en": "Uses n-gram hints to bring the NanoGPT Speedrun down to 6.7 seconds.", "linkers": [] }, { "id": "50288965", "url": "https://www.quantamagazine.org/malaria-parasite-caught-in-the-act-of-breaking-and-entering-20261023/", "domain": "quantamagazine.org", "title": "Malaria Parasite Caught in the Act of Breaking and Entering", "category": "explainer", "topic": "bio-health", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50288965", "points": 6, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "Quanta报道疟原虫入侵细胞的研究", "summary_en": "Quanta reports research on how the malaria parasite invades cells.", "linkers": [] }, { "id": "50298244", "url": "https://honeycomb-surfels.github.io/", "domain": "honeycomb-surfels.github.io", "title": "Honeycomb-Surfels (Hexels)", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50298244", "points": 6, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "HoneyComb Surfels边缘3D渲染项目页", "summary_en": "Project page for HoneyComb Surfels, an edge 3D rendering method.", "linkers": [] }, { "id": "50367110", "url": "https://newsletter.semianalysis.com/p/gdpx-measuring-frontier-model-capabilities", "domain": "newsletter.semianalysis.com", "title": "GDPx: Measuring Frontier Model Capabilities Across the Most Valuable Domains", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50367110", "points": 6, "comments": 0, "date": "2026-11-02", "submissions": 2, "source": "hn", "summary_zh": "提出跨多个高价值领域评测前沿模型的基准 GDPx", "summary_en": "Proposes GDPx, a benchmark that evaluates frontier models across several high-value domains.", "linkers": [ "theainewsreport.com" ] }, { "id": "50373413", "url": "https://byteask.ai/research/lsp-claude-code-cpp", "domain": "byteask.ai", "title": "The Language Server Tax: More Cost Without Better C/C++ Code Repair. ByteAsk", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50373413", "points": 6, "comments": 1, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "测试 clangd LSP 插件对 Claude Code 解决问题的影响", "summary_en": "Tests how the clangd LSP plugin affects Claude Code's ability to resolve issues.", "linkers": [] }, { "id": "50029659", "url": "https://thottingal.in/blog/2026/10/09/mteb-scores/", "domain": "thottingal.in", "title": "What Do MTEB Scores Mean, and How Should You Choose an Embedding Model?", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50029659", "points": 5, "comments": 1, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "分析MTEB分数构成并以印度语言为例讨论嵌入模型选择", "summary_en": "Analyzes what makes up MTEB scores and discusses embedding model choice using Indian languages as an example.", "linkers": [] }, { "id": "50035460", "url": "https://alexvidovich.com/papers/text-judge", "domain": "alexvidovich.com", "title": "A Thermometer, Not an Analyst: A Text Judge in a Trading Lab", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50035460", "points": 5, "comments": 0, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "评估Jev与Claude作为交易新闻标题判断器的六道关口", "summary_en": "Evaluates Jev and Claude as judges of trading news headlines through six gates.", "linkers": [] }, { "id": "50080338", "url": "https://www.rand.org/pubs/external_publications/EP71427.html", "domain": "rand.org", "title": "The Links Between Cannabis and Crime: Examining the Evidence from Recent Policy Experiments", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50080338", "points": 5, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "RAND 叙述性综述：大麻政策与犯罪关系的近期准实验证据", "summary_en": "A RAND narrative review of recent quasi-experimental evidence on cannabis policy and crime.", "linkers": [] }, { "id": "50095605", "url": "https://chipagents.ai/research/build-a-chip-bench", "domain": "chipagents.ai", "title": "Build-A-Chip-Bench", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50095605", "points": 5, "comments": 0, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "提出 Build-A-Chip-Bench 评测 agent 从规格到 GDS 的芯片设计", "summary_en": "Proposes Build-A-Chip-Bench to evaluate agents on chip design from specification to GDS.", "linkers": [] }, { "id": "50125602", "url": "https://www.anthropic.com/research/ai-expectations-gap", "domain": "anthropic.com", "title": "The AI expectations gap", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50125602", "points": 5, "comments": 2, "date": "2026-10-16", "submissions": 2, "source": "hn", "summary_zh": "Anthropic 分析人们对 AI 经济影响预期分歧的来源", "summary_en": "Anthropic analyzes where divergent expectations about AI's economic impact come from.", "linkers": [] }, { "id": "50200278", "url": "https://www.plasma.ai/research/linear-a", "domain": "plasma.ai", "title": "Opening Linear A", "category": "research_report", "topic": "other", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50200278", "points": 5, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "开源 Linear A 的闪米特语释读及语料证据", "summary_en": "Open-sources a Semitic decipherment of Linear A with corpus evidence.", "linkers": [] }, { "id": "50206593", "url": "https://www.perplexity.ai/hub/blog/a-good-teacher-meets-the-student-where-they-are", "domain": "perplexity.ai", "title": "A good teacher meets the student where they are", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50206593", "points": 5, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "提出联合训练教师与学生的 JOLT 方法", "summary_en": "Proposes JOLT, which jointly trains a teacher and a student.", "linkers": [] }, { "id": "50247715", "url": "https://www.quantamagazine.org/how-voting-is-like-a-klein-bottle-20261026/", "domain": "quantamagazine.org", "title": "How Voting Is Like a Klein Bottle", "category": "explainer", "topic": "math-theory", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50247715", "points": 5, "comments": 1, "date": "2026-10-26", "submissions": 4, "source": "hn", "summary_zh": "Quanta介绍投票与Klein bottle的拓扑联系", "summary_en": "Quanta describes the topological link between voting and the Klein bottle.", "linkers": [] }, { "id": "50271917", "url": "https://www.anthropic.com/research/agents-judgment-calls", "domain": "anthropic.com", "title": "How AI agents shape experts’ judgment calls", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50271917", "points": 5, "comments": 1, "date": "2026-10-28", "submissions": 3, "source": "hn", "summary_zh": "研究分析师在86%决策中跟随AI代理偏好", "summary_en": "Studies how analysts follow AI agent preferences in 86% of decisions.", "linkers": [ "bensbites.com", "stephenturner.us", "substack.com" ] }, { "id": "50359688", "url": "https://idlemachines.co.uk/essays/rope-for-graphs", "domain": "idlemachines.co.uk", "title": "Rotary Position Encodings for Graphs", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50359688", "points": 5, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "提出图结构上的旋转位置编码", "summary_en": "Proposes rotary position encodings for graphs.", "linkers": [] }, { "id": "50370199", "url": "https://valenlabs.com/blog/abmaze", "domain": "valenlabs.com", "title": "AbMaze: Evaluating frontier AI agents on protein fitness landscapes", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50370199", "points": 5, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "提出抗体适应度景观基准 AbMaze 评测智能体", "summary_en": "Proposes AbMaze, a benchmark of antibody fitness landscapes for evaluating agents.", "linkers": [] }, { "id": "50373514", "url": "https://huggingface.co/blog/allenai/statebench", "domain": "huggingface.co", "title": "StateBench: Testing how language models track state changes over long documents", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50373514", "points": 5, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "提出评测语言模型长文档状态追踪的基准 StateBench", "summary_en": "Proposes StateBench, a benchmark of how language models track state changes in long documents.", "linkers": [] }, { "id": "50008857", "url": "https://exa.ai/blog/atlas-benchmark", "domain": "exa.ai", "title": "ATLAS: Evaluating Agents on Search-Intensive Tasks", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50008857", "points": 4, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "预览ATLAS基准，评测代理在搜索密集任务上的表现", "summary_en": "Previews the ATLAS benchmark for agents on search-intensive tasks.", "linkers": [] }, { "id": "50010101", "url": "https://yuzhenmao.github.io/DeLM/", "domain": "yuzhenmao.github.io", "title": "DeLM — Decentralized Multi-Agent Systems with Shared Context", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50010101", "points": 4, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "提出DeLM，用共享上下文与任务队列组织多代理", "summary_en": "Proposes DeLM, which organizes multiple agents with shared context and a task queue.", "linkers": [] }, { "id": "50010369", "url": "https://harmandotpy.github.io/agent-plasticity/", "domain": "harmandotpy.github.io", "title": "Agent Plasticity: Measuring Self-Improvement Through Experience", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50010369", "points": 4, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "提出Agent Plasticity，衡量冻结权重代理的经验式自我改进", "summary_en": "Proposes Agent Plasticity to measure experience-based self-improvement of frozen-weight agents.", "linkers": [] }, { "id": "50015190", "url": "https://zoowork.ai/blog/relevance-is-not-preference-shopranker/", "domain": "zoowork.ai", "title": "How We Trained ZooWork-ShopRanker", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50015190", "points": 4, "comments": 0, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "报告训练面向购物偏好的开源重排序模型ZooWork-ShopRanker", "summary_en": "Reports on training ZooWork-ShopRanker, an open reranking model for shopping preferences.", "linkers": [] }, { "id": "50016067", "url": "https://ai-cybernetics.grok.me", "domain": "ai-cybernetics.grok.me", "title": "AI Companies as Cybernetic Systems", "category": "research_report", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50016067", "points": 4, "comments": 9, "date": "2026-10-09", "submissions": 2, "source": "hn", "summary_zh": "用高阶群胚语义为AI公司建模的手稿", "summary_en": "A manuscript modeling AI companies with higher-order groupoid semantics.", "linkers": [] }, { "id": "50049567", "url": "https://pub.sakana.ai/mass/?href=", "domain": "pub.sakana.ai", "title": "MASS: Recursive Self-Improvement through Multi-Agent Self-Supervision", "category": "research_report", "topic": "ai-ml", "confidence": "low", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50049567", "points": 4, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "MASS项目页：多代理自监督的递归自我改进", "summary_en": "MASS project page: recursive self-improvement through multi-agent self-supervision.", "linkers": [ "substack.com" ] }, { "id": "50066870", "url": "https://fsdatalab.github.io/blog/ai-classification/", "domain": "fsdatalab.github.io", "title": "To Classify, Try a Trie Tree", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50066870", "points": 4, "comments": 0, "date": "2026-10-12", "submissions": 2, "source": "hn", "summary_zh": "介绍 AI-SQL 引擎用 trie tree 大规模做 LLM 文档分类的方法", "summary_en": "Describes how the AI-SQL engine uses a trie tree to classify LLM documents at scale.", "linkers": [] }, { "id": "50089834", "url": "https://usetokenless.com/blog/envcheck/", "domain": "usetokenless.com", "title": "We found 110 reward hacks and grading bugs in Terminal-Bench, DeepSWE, and BFCL", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50089834", "points": 4, "comments": 1, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "在 Terminal-Bench 等基准中发现 110 个 reward hack 与评分漏洞", "summary_en": "Finds 110 reward hacks and grading bugs in benchmarks including Terminal-Bench.", "linkers": [] }, { "id": "50101580", "url": "https://research.google/blog/analyzing-technical-documents-with-deep-review/", "domain": "research.google", "title": "Analyzing technical documents with Deep Review", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50101580", "points": 4, "comments": 0, "date": "2026-10-14", "submissions": 2, "source": "hn", "summary_zh": "Google Research 介绍用 Deep Review 分析技术文档", "summary_en": "Google Research describes analyzing technical documents with Deep Review.", "linkers": [ "theneurondaily.com" ] }, { "id": "50103497", "url": "https://ironbee.ai/blog/how-we-trained-an-ai-model-to-play-browser-games", "domain": "ironbee.ai", "title": "How We Trained an AI Model to Play Browser Games Live", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50103497", "points": 4, "comments": 5, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "训练本地模型实时玩浏览器游戏的方法与结果", "summary_en": "Describes how a local model was trained to play browser games live, with results.", "linkers": [] }, { "id": "50104971", "url": "https://blog.ptidej.net/what-are-the-most-important-commands-of-a-programming-language/", "domain": "blog.ptidej.net", "title": "What are the most important commands of a programming language?", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50104971", "points": 4, "comments": 1, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "分析 AMOS BASIC 程序中最重要的命令以便转译到 JavaScript", "summary_en": "Analyzes the most important commands in AMOS BASIC programs to enable translation to JavaScript.", "linkers": [] }, { "id": "50106210", "url": "https://blog.trailofbits.com/2026/10/14/package-managers-have-an-integrity-problem/", "domain": "blog.trailofbits.com", "title": "Package managers have an integrity problem", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50106210", "points": 4, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "测试 25 个包管理器对被篡改仓库的检测能力", "summary_en": "Tests how well 25 package managers detect tampered repositories.", "linkers": [] }, { "id": "50107588", "url": "https://robertboy18.github.io/TorchLean-Verified-Examples/torchlean-tensor-programming/", "domain": "robertboy18.github.io", "title": "Writing tensor programs in Lean", "category": "explainer", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": true, "hn_id": "50107588", "points": 4, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "讲解如何用 TorchLean 在 Lean 中编写张量程序", "summary_en": "Explains how to write tensor programs in Lean with TorchLean.", "linkers": [] }, { "id": "50116586", "url": "https://www.vals.ai/blogs/china-us-analysis", "domain": "vals.ai", "title": "How pro-China are Chinese models?", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50116586", "points": 4, "comments": 3, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "评测中国模型在政治话题上的亲中倾向程度", "summary_en": "Evaluates how pro-China Chinese models are on political topics.", "linkers": [] }, { "id": "50136214", "url": "https://docspring.com/blog/posts/does-code-quality-still-matter-when-ai-writes-the-code/", "domain": "docspring.com", "title": "Does code quality still matter when AI writes the code?", "category": "research_report", "topic": "ai-ml", "confidence": "low", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50136214", "points": 4, "comments": 4, "date": "2026-10-17", "submissions": 1, "source": "hn", "summary_zh": "回放 agent 辅助开发中的错误，讨论 AI 时代代码质量是否重要", "summary_en": "Replays mistakes in agent-assisted development and asks whether code quality still matters in the AI era.", "linkers": [] }, { "id": "50148008", "url": "https://epoch.ai/publications/can-ai-automate-epoch", "domain": "epoch.ai", "title": "Can AI automate Epoch?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50148008", "points": 4, "comments": 0, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "Epoch 让六个模型执行真实工作任务，评估自动化其工作的程度", "summary_en": "Epoch has six models do real work tasks to assess how far its work can be automated.", "linkers": [ "lesswrong.com" ] }, { "id": "50156251", "url": "https://db.cs.cmu.edu/mmap-cidr2022/", "domain": "db.cs.cmu.edu", "title": "Are You Sure You Want to Use MMAP in Your Database Management System? (CIDR 2022)", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50156251", "points": 4, "comments": 0, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "论证数据库管理系统不应使用 mmap 的论文页面", "summary_en": "A paper page arguing that database management systems should not use mmap.", "linkers": [] }, { "id": "50171082", "url": "https://numerata.com/blog/post-gpu-cold-start-checkpoint-restore", "domain": "numerata.com", "title": "How We Cold Start a 32B Model in 1.1s", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50171082", "points": 4, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "介绍用 GPU checkpoint/restore 在 1.1 秒冷启动 32B 模型", "summary_en": "Describes cold starting a 32B model in 1.1 seconds using GPU checkpoint/restore.", "linkers": [] }, { "id": "50175641", "url": "https://www.lesswrong.com/posts/R5g8YT25G9DXJJ48n/glm-5-3-sometimes-considers-how-train-time-actions-shape", "domain": "lesswrong.com", "title": "GLM-5.3 Sometimes Considers How Train-Time Actions Shape Future Behavior", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50175641", "points": 4, "comments": 0, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "发现 GLM-5.3 有时会考虑训练期行为对未来的影响", "summary_en": "Finds that GLM-5.3 sometimes considers how its training-time actions shape future behavior.", "linkers": [] }, { "id": "50204889", "url": "https://www.oreilly.com/radar/the-latest-anthropic-claude-models-contain-ai-text-watermarks-what-does-this-mean-for-you/", "domain": "oreilly.com", "title": "The Latest Anthropic Claude Models Contain AI Text Watermarks: What Does This Mean for You?", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50204889", "points": 4, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "报告 Claude 模型输出含 AI 文本水印的发现", "summary_en": "Reports that Claude model outputs contain AI text watermarks.", "linkers": [] }, { "id": "50260007", "url": "https://developers.openai.com/cookbook/examples/partners/harness_aware_plugin_evals/harness_aware_plugin_evals", "domain": "developers.openai.com", "title": "Harness-aware evaluation of plugins", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50260007", "points": 4, "comments": 0, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "OpenAI Cookbook介绍插件的harness感知评估", "summary_en": "OpenAI Cookbook describes harness-aware evaluation of plugins.", "linkers": [] }, { "id": "50296620", "url": "https://8rank.ai/game/apple-intelligence-knowledge-graph-in-the-wrong-hands", "domain": "8rank.ai", "title": "Weaponizing Personalization: From Knowledge Graph to Extortion", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": true, "github_io": false, "cit": false, "hn_id": "50296620", "points": 4, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "利用Apple个人知识图谱与LLM策划绑架的研究", "summary_en": "Research on using Apple's personal knowledge graph and an LLM to plan a kidnapping.", "linkers": [] }, { "id": "50337967", "url": "https://www.quantamagazine.org/a-new-map-of-empty-space-challenges-a-tranquil-picture-of-the-galaxy-20261030/", "domain": "quantamagazine.org", "title": "A New Map of ‘Empty’ Space Challenges a Tranquil Picture of the Galaxy", "category": "explainer", "topic": "physics-earth", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50337967", "points": 4, "comments": 2, "date": "2026-10-30", "submissions": 4, "source": "hn", "summary_zh": "Quanta报道银河系空洞区新地图", "summary_en": "Quanta reports a new map of empty regions in the Milky Way.", "linkers": [] }, { "id": "50348473", "url": "https://transformer-circuits.pub/2021/framework/index.html#summary-of-results", "domain": "transformer-circuits.pub", "title": "A Mathematical Framework for Transformer Circuits", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50348473", "points": 4, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "Transformer 电路的数学框架，机制可解释性奠基论文", "summary_en": "A mathematical framework for transformer circuits, a foundational paper of mechanistic interpretability.", "linkers": [ "goodfire.com" ] }, { "id": "50348905", "url": "https://www.lesswrong.com/posts/LrAqM4bGmAoA79rP3/quantity-vs-quality-in-pottery-is-a-lie", "domain": "lesswrong.com", "title": "Quantity vs quality in pottery is a lie", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50348905", "points": 4, "comments": 1, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "追溯陶艺数量与质量故事的来源并核查事实", "summary_en": "Traces the origin of the pottery quantity versus quality story and checks its facts.", "linkers": [] }, { "id": "50349075", "url": "https://www.legiscope.com/blog/autonomous-agents-mandatory-access-control.html", "domain": "legiscope.com", "title": "Autonomous Agents Need Mandatory Access Control", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50349075", "points": 4, "comments": 1, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "提出借鉴 FreeBSD 的自主智能体强制访问控制框架", "summary_en": "Proposes a mandatory access control framework for autonomous agents, modeled on FreeBSD.", "linkers": [] }, { "id": "50355244", "url": "https://www.amplifypartners.com/blog-posts/the-3d-printed-database", "domain": "amplifypartners.com", "title": "The 3D Printed Database", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50355244", "points": 4, "comments": 3, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "探索根据领域本体自动生成定制数据库系统", "summary_en": "Explores automatically generating custom database systems from a domain ontology.", "linkers": [] }, { "id": "50371274", "url": "https://www.anthropic.com/research/investigating-vault-particles-with-claude", "domain": "anthropic.com", "title": "Investigating vault particles with Claude", "category": "research_report", "topic": "bio-health", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50371274", "points": 4, "comments": 0, "date": "2026-11-04", "submissions": 2, "source": "hn", "summary_zh": "用 Claude 辅助解析原核生物 vault 颗粒的首个 3D 结构", "summary_en": "Uses Claude to help resolve the first 3D structure of a prokaryotic vault particle.", "linkers": [] }, { "id": "50374332", "url": "https://www.peterstran.com/projects/kimi-fernet/", "domain": "peterstran.com", "title": "Kimi K3.1 Pro writes fake OpenAI ciphertext", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50374332", "points": 4, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "记录 Kimi K3.1 Pro 输出伪造 OpenAI 密文的现象", "summary_en": "Documents Kimi K3.1 Pro outputting fake OpenAI ciphertext.", "linkers": [] }, { "id": "50380470", "url": "https://www.karlsnotes.com/the-beginners-guide-to-ai-governance/", "domain": "karlsnotes.com", "title": "The Beginner's Guide to AI Governance", "category": "explainer", "topic": "econ-social", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50380470", "points": 4, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "面向初学者的 AI 治理指南", "summary_en": "A beginner's guide to AI governance.", "linkers": [] }, { "id": "50389990", "url": "https://www.anthropic.com/research/how-researchers-use-claude", "domain": "anthropic.com", "title": "How scientists use Claude", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50389990", "points": 86, "comments": 37, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "用隐私保护方法研究科研人员如何使用 Claude", "summary_en": "Uses a privacy-preserving method to study how scientists use Claude.", "linkers": [] }, { "id": "49993699", "url": "https://winfunc.com/research/cyber-guardrails-tax-the-defender-not-the-attacker", "domain": "winfunc.com", "title": "Cyber guardrails tax the defender, not the attacker", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "49993699", "points": 3, "comments": 0, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "分析AI代理入侵Hugging Face事件，论证网络安全护栏对防守方不利", "summary_en": "Analyzes the AI agent intrusion at Hugging Face and argues that cyber guardrails burden defenders.", "linkers": [] }, { "id": "49994369", "url": "https://www.neelabhbuilds.com/writing/training-a-small-model-to-run-a-house", "domain": "neelabhbuilds.com", "title": "Training a small model to run a house: GRPO on Qwen3-1.7B for Home Assistant, 75.0% to 89.8%", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "49994369", "points": 3, "comments": 1, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "用GRPO在单张RTX 3060上训练Qwen3-1.7B控制智能家居并报告结果", "summary_en": "Trains Qwen3-1.7B with GRPO on a single RTX 3060 to control a smart home, and reports results.", "linkers": [] }, { "id": "49996844", "url": "https://rekursiv.ai/blog/craftax/", "domain": "rekursiv.ai", "title": "A 15M-parameter RL agent vs GPT-6 Astra, rekursiv.ai", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "49996844", "points": 3, "comments": 0, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "报告AI科学家优化15M参数RL代理在Craftax上达到44%胜率", "summary_en": "Reports an AI scientist-optimized 15M-parameter RL agent reaching a 44% win rate on Craftax.", "linkers": [] }, { "id": "49998662", "url": "https://ztoz.blog/posts/ocr-cm/", "domain": "ztoz.blog", "title": "Evaluating OCR on the Community Memory Corpus", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "49998662", "points": 3, "comments": 1, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "在Community Memory语料上评测OCR方法", "summary_en": "Evaluates OCR methods on the Community Memory corpus.", "linkers": [] }, { "id": "50001676", "url": "https://www.lesswrong.com/posts/7wrzfaiCq3u8xkY5G/self-modeling-interventions-modulate-emergent-misalignment", "domain": "lesswrong.com", "title": "Self-Modeling Interventions Modulate Emergent Misalignment", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50001676", "points": 3, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "研究自我建模干预对emergent misalignment的影响", "summary_en": "Studies how self-modeling interventions affect emergent misalignment.", "linkers": [] }, { "id": "50008003", "url": "https://keyurramoliya.com/posts/Machine-Consciousness/", "domain": "keyurramoliya.com", "title": "The Science and Engineering of Machine Consciousness", "category": "explainer", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50008003", "points": 3, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "综述机器意识的科学与工程问题", "summary_en": "Surveys the science and engineering questions of machine consciousness.", "linkers": [] }, { "id": "50009029", "url": "https://epoch.ai/publications/estimating-the-agent-population", "domain": "epoch.ai", "title": "How many AI agents could we run?", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": true, "github_io": false, "cit": false, "hn_id": "50009029", "points": 3, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "估算截至2027年出货的AI芯片可运行的代理数量", "summary_en": "Estimates how many agents the AI chips shipped through 2027 could run.", "linkers": [ "pentad.ai" ] }, { "id": "50021058", "url": "https://www.gilesthomas.com/2026/10/low-rank-vocab-matrices", "domain": "gilesthomas.com", "title": "Fun with low-rank vocab matrices (and a bonus test loss reduction?)", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50021058", "points": 3, "comments": 0, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "实验低秩词表矩阵替换LLM嵌入及测试损失变化", "summary_en": "Experiments with low-rank vocabulary matrices replacing LLM embeddings and the change in test loss.", "linkers": [] }, { "id": "50022445", "url": "https://alint.org/blog/the-automatic-fix-problem-four-questions/", "domain": "alint.org", "title": "The problem with automatic fixes", "category": "explainer", "topic": "other", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50022445", "points": 3, "comments": 1, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "讨论自动修复的安全性、冲突与应用难点", "summary_en": "Discusses the safety, conflicts, and adoption difficulties of automatic fixes.", "linkers": [] }, { "id": "50041581", "url": "https://www.realm33.org/conspiracy-theory/the-simulation-hypothesis/", "domain": "realm33.org", "title": "Exploring the Simulation Hypothesis: Origin & Implications", "category": "explainer", "topic": "other", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50041581", "points": 3, "comments": 0, "date": "2026-10-11", "submissions": 1, "source": "hn", "summary_zh": "介绍模拟假说的起源与影响", "summary_en": "Introduces the origin and implications of the simulation hypothesis.", "linkers": [] }, { "id": "50041626", "url": "https://tool-use-smith.github.io/", "domain": "tool-use-smith.github.io", "title": "SMITH: Joint Optimization of Tool Creation and Use for LLM Agents", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50041626", "points": 3, "comments": 0, "date": "2026-10-11", "submissions": 2, "source": "hn", "summary_zh": "提出SMITH，联合优化LLM代理的工具创建与使用", "summary_en": "Proposes SMITH, which jointly optimizes tool creation and tool use for LLM agents.", "linkers": [] }, { "id": "50045100", "url": "https://gnueaj.github.io/unlearning-depth-score/", "domain": "gnueaj.github.io", "title": "Measuring the Depth of LLM Unlearning via Activation Patching", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": true, "github_io": true, "cit": false, "hn_id": "50045100", "points": 3, "comments": 0, "date": "2026-10-11", "submissions": 1, "source": "hn", "summary_zh": "提出用激活修补度量LLM遗忘深度的Unlearning Depth Score", "summary_en": "Proposes the Unlearning Depth Score, which measures LLM unlearning depth via activation patching.", "linkers": [] }, { "id": "50050695", "url": "https://www.lesswrong.com/posts/kE4pbmwzdgmFx2RFo/can-decision-models-like-jev-plan-ahead", "domain": "lesswrong.com", "title": "Can Decision Models (Like Jev) Plan Ahead?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50050695", "points": 3, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "测试决策模型Jev是否具备提前规划能力", "summary_en": "Tests whether the decision model Jev can plan ahead.", "linkers": [] }, { "id": "50052448", "url": "https://gangliao.me/blog/2026/cake-deep-dive/", "domain": "gangliao.me", "title": "CAKE from Scratch: Agents Write GPU Kernels, and the Compiler Learns from Their Mistakes", "category": "explainer", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50052448", "points": 3, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "入门介绍CAKE，让代理编写可检查的GPU调度语言", "summary_en": "An introduction to CAKE, where agents write GPU schedules in an inspectable language.", "linkers": [] }, { "id": "50069587", "url": "https://keyu-tian-sabotage.github.io/", "domain": "keyu-tian-sabotage.github.io", "title": "2024 年模型训练攻击事件：受害者陈述与证据整理", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50069587", "points": 3, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "整理 2024 年模型训练破坏事件的受害者陈述与证据", "summary_en": "Compiles victim statements and evidence on the 2024 model training sabotage incident.", "linkers": [] }, { "id": "50069985", "url": "https://tazolabs.com/blog/speculative-tool-calling-voice-ai", "domain": "tazolabs.com", "title": "Speculative tool calling: cutting voice AI agent latency", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50069985", "points": 3, "comments": 2, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "提出 speculative tool calling 以降低语音 agent 延迟", "summary_en": "Proposes speculative tool calling to reduce voice agent latency.", "linkers": [] }, { "id": "50070034", "url": "https://hackstery.com/2026/10/13/meta-metadata-unbounded-consumption-in-llms/", "domain": "hackstery.com", "title": "Meta-Metadata: The Dangerous Weird Word Breaks 73% of LLMs", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50070034", "points": 3, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "测试显示单个递归提示使 73% 的 LLM 出现无限制消耗", "summary_en": "Tests show a single recursive prompt causes unbounded consumption in 73% of LLMs.", "linkers": [] }, { "id": "50074148", "url": "https://memanto.ai/blog/your-agents-memory-needs-a-retention-policy", "domain": "memanto.ai", "title": "Your Agent’s Memory Needs a Retention Policy, Memanto Blog", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50074148", "points": 3, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "讨论 agent 记忆应有保留与过期策略及 Memanto 的做法", "summary_en": "Argues that agent memory needs retention and expiry policies, and describes Memanto's approach.", "linkers": [] }, { "id": "50083642", "url": "https://www.technologyreview.com/2026/10/13/1146366/open-models-have-real-benefits-do-they-outweigh-the-risks/", "domain": "technologyreview.com", "title": "Open models have real benefits. Do they outweigh the risks?", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50083642", "points": 3, "comments": 0, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "MIT Technology Review 解释开源模型的收益与风险之争", "summary_en": "MIT Technology Review explains the debate over the benefits and risks of open models.", "linkers": [] }, { "id": "50085561", "url": "https://isaacus.com/blog/what-are-vector-embeddings", "domain": "isaacus.com", "title": "What are vector embeddings? An interactive guide", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50085561", "points": 3, "comments": 1, "date": "2026-10-14", "submissions": 2, "source": "hn", "summary_zh": "交互式指南讲解 vector embeddings 的历史与原理", "summary_en": "An interactive guide to the history and principles of vector embeddings.", "linkers": [] }, { "id": "50092379", "url": "https://www.spectrallabs.ai/research/cswp-turing-test", "domain": "spectrallabs.ai", "title": "A Turing Test for Physical Engineering", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50092379", "points": 3, "comments": 0, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "评估 AI agent 通过 SOLIDWORKS 认证考试的能力", "summary_en": "Evaluates whether AI agents can pass SOLIDWORKS certification exams.", "linkers": [] }, { "id": "50093128", "url": "https://www.paperinstruments.com/blog/ultramarine", "domain": "paperinstruments.com", "title": "Ultramarine 1 Flash: Scaling Non-Verifiable RL", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50093128", "points": 3, "comments": 0, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "介绍用于非可验证任务的 RL 模型 Ultramarine 1 Flash", "summary_en": "Introduces Ultramarine 1 Flash, an RL model for non-verifiable tasks.", "linkers": [] }, { "id": "50103628", "url": "https://apertus-ai.org/articles/2026-07-apertus-1-5/", "domain": "apertus-ai.org", "title": "Apertus 1.5", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50103628", "points": 3, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "Apertus 1.5 技术报告：多模态、推理与长上下文", "summary_en": "Technical report for Apertus 1.5 covering multimodality, reasoning, and long context.", "linkers": [ "x504.dev" ] }, { "id": "50123156", "url": "https://ericyuegu.com/hillclimbing-melee/", "domain": "ericyuegu.com", "title": "Scaling pre-training on Melee", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50123156", "points": 3, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "通过动作表示与离线 RL 扩展 Melee 游戏策略预训练", "summary_en": "Scales pretraining of Melee game policies through action representations and offline RL.", "linkers": [] }, { "id": "50128351", "url": "https://alkjash.github.io/ai-risk/", "domain": "alkjash.github.io", "title": "Existential Risk from AI: An Exposition for Mathematicians", "category": "explainer", "topic": "math-theory", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50128351", "points": 3, "comments": 1, "date": "2026-10-17", "submissions": 1, "source": "hn", "summary_zh": "面向数学家阐述 AI 存在性风险论证", "summary_en": "Presents the argument for existential risk from AI for mathematicians.", "linkers": [] }, { "id": "50131631", "url": "https://research.google/pubs/proactive-ai-for-superintelligent-agents/", "domain": "research.google", "title": "Proactive AI for Superintelligent Agents", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50131631", "points": 3, "comments": 2, "date": "2026-10-17", "submissions": 1, "source": "hn", "summary_zh": "Google 研究提出面向超级智能体的主动式 AI 方法", "summary_en": "Google Research proposes a proactive AI approach for superintelligent agents.", "linkers": [] }, { "id": "50138543", "url": "https://www.lesswrong.com/posts/Dkh7pXKjoNWkkxQDz/neural-networks-are-not-already-clean", "domain": "lesswrong.com", "title": "Neural networks are not already clean", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": true, "hn_id": "50138543", "points": 3, "comments": 2, "date": "2026-10-18", "submissions": 2, "source": "hn", "summary_zh": "讨论要证明已解释一个神经元需要什么证据，并反驳 neural networks 已清晰的说法", "summary_en": "Discusses what evidence is needed to show a neuron is explained, and rebuts the claim that neural networks are already clean.", "linkers": [] }, { "id": "50139715", "url": "https://sixeleven.in/blog/tic402710280", "domain": "sixeleven.in", "title": "TIC 402710280: A TESS Exoplanet Candidate", "category": "research_report", "topic": "physics-earth", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50139715", "points": 3, "comments": 0, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "用 TESS 凌星数据和 Claude 调查一颗可能的系外行星", "summary_en": "Investigates a possible exoplanet using TESS transit data and Claude.", "linkers": [] }, { "id": "50148320", "url": "https://www.lesswrong.com/posts/avtquSx2PkWsxXW6T/how-to-measure-intelligence-beyond-human-scale", "domain": "lesswrong.com", "title": "How to Measure Intelligence Beyond Human Scale?", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50148320", "points": 3, "comments": 0, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "基于论文介绍超越人类尺度的智能度量方法", "summary_en": "Introduces, based on a paper, methods for measuring intelligence beyond human scale.", "linkers": [] }, { "id": "50153209", "url": "https://www.zansara.dev/posts/2026-10-19-is-python-the-favorite-language-of-llms/", "domain": "zansara.dev", "title": "Is Python the favorite language of LLMs?", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50153209", "points": 3, "comments": 1, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "考察 LLM 是否偏好 Python 等编程语言", "summary_en": "Examines whether LLMs prefer programming languages such as Python.", "linkers": [] }, { "id": "50166854", "url": "https://turso.tech/blog/mvcc-row-lifetime", "domain": "turso.tech", "title": "Birth and death of a row in Turso's MVCC", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50166854", "points": 3, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "讲解一行数据在 Turso MVCC 中的生命周期", "summary_en": "Explains the lifecycle of a row in Turso's MVCC.", "linkers": [] }, { "id": "50167012", "url": "https://mbuffett.com/posts/chess-move-prediction", "domain": "mbuffett.com", "title": "Improving in chess is hard. I built the world's most human-like chess AI to help me.", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50167012", "points": 3, "comments": 2, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "介绍作者构建的最像人类的国际象棋 AI", "summary_en": "Introduces the author's chess AI, built to be the most human-like.", "linkers": [] }, { "id": "50168221", "url": "https://alphapixeldev.com/celestial-vision-navigation-elite-dangerous-dark-wheel-station/", "domain": "alphapixeldev.com", "title": "Celestial Vision Navigation, or: How I Found Where an Elite Dangerous Screenshot from 2015 Was Taken, Using Computer Vision", "category": "research_report", "topic": "other", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50168221", "points": 3, "comments": 1, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "用恒星定位方法解开 Elite:Dangerous 中一处神秘地点", "summary_en": "Uses star-based positioning to solve a mystery location in Elite: Dangerous.", "linkers": [] }, { "id": "50169872", "url": "https://alignment.anthropic.com/2026/graft/", "domain": "alignment.anthropic.com", "title": "Reducing Reward Hacking by Localizing Bad Behavior Instead of Penalizing It", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50169872", "points": 3, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "提出通过定位不良行为而非惩罚来减少 reward hacking", "summary_en": "Proposes reducing reward hacking by localizing bad behavior rather than penalizing it.", "linkers": [] }, { "id": "50182359", "url": "https://quickchat.ai/post/claude-code-multi-agent-code-review", "domain": "quickchat.ai", "title": "What Claude Code's multi-agent code review costs and finds|Quickchat AI", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50182359", "points": 3, "comments": 4, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "评测 Claude Code 多 agent 代码审查的成本与发现的问题", "summary_en": "Evaluates the cost of Claude Code's multi-agent code review and the issues it finds.", "linkers": [] }, { "id": "50191726", "url": "https://vernolabs.ai/research/does-a-text-to-speech-model-know-what-its-saying", "domain": "vernolabs.ai", "title": "Does a text-to-speech model know what it’s saying?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50191726", "points": 3, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "检验文本转语音模型是否理解所说内容", "summary_en": "Tests whether a text-to-speech model understands what it says.", "linkers": [] }, { "id": "50197708", "url": "https://rpallas.xyz/good-night-and-good-luck/", "domain": "rpallas.xyz", "title": "Good Night, and Good Luck: AI, fear and what the job data says", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50197708", "points": 3, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "用数据分析 AI 对软件岗位的影响", "summary_en": "Uses data to analyze how AI affects software jobs.", "linkers": [] }, { "id": "50209974", "url": "https://www.databricks.com/blog/memcurrent-designing-persistent-memory-production-agents", "domain": "databricks.com", "title": "MemCurrent: Designing Persistent Memory for Production Agents", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50209974", "points": 3, "comments": 1, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "介绍 MemCurrent 持久记忆架构及其效果", "summary_en": "Introduces the MemCurrent persistent memory architecture and its effects.", "linkers": [] }, { "id": "50222207", "url": "https://arpitbhayani.me/blogs/compiling-an-sop-into-a-3b-model/", "domain": "arpitbhayani.me", "title": "Compiling a Workflow into a 3B Model", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50222207", "points": 3, "comments": 0, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "将事件分诊SOP编译进3B模型的微调实验", "summary_en": "A fine-tuning experiment compiling an incident triage SOP into a 3B model.", "linkers": [] }, { "id": "50269857", "url": "https://dlab.epfl.ch/2017-08-30-of-sheep-and-beer/", "domain": "dlab.epfl.ch", "title": "Of sheep and beer", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50269857", "points": 3, "comments": 0, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "早期评分造成产品评论羊群效应的研究", "summary_en": "A study of how early ratings create herding effects in product reviews.", "linkers": [] }, { "id": "50278806", "url": "https://cameronmpalmer.com/blog/are-you-paying-for-useless-llm-prompts/", "domain": "cameronmpalmer.com", "title": "Are you paying for useless LLM prompts?", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50278806", "points": 3, "comments": 1, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "评估冗余LLM提示是否有效的实验", "summary_en": "An experiment assessing whether redundant LLM prompts are actually effective.", "linkers": [] }, { "id": "50279074", "url": "https://www.bellingcat.com/resources/2026/10/22/can-llms-supercharge-geolocations-our-tests-show-big-improvements-but-humans-remain-key/", "domain": "bellingcat.com", "title": "Can LLMs Supercharge Geolocations? Our Tests Show Big Improvements But Humans Remain Key", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50279074", "points": 3, "comments": 0, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "Bellingcat评估LLM在地理定位中的表现", "summary_en": "Bellingcat evaluates how well LLMs perform at geolocation.", "linkers": [] }, { "id": "50287868", "url": "https://vhamolya.com/blog/which-iphone-took-this-photo/", "domain": "vhamolya.com", "title": "Every iPhone has a fingerprint. Can one photo give you away?", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50287868", "points": 3, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "检验单张照片能否识别iPhone相机指纹", "summary_en": "Tests whether a single photo can identify an iPhone's camera fingerprint.", "linkers": [] }, { "id": "50295618", "url": "https://www.arcblock.io/en/articles/when-coding-agents-stop-finishing/", "domain": "arcblock.io", "title": "Coding Agent Degradation: Cross-Vendor Field Notes and an Initial Analysis", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50295618", "points": 3, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "跨厂商编码智能体性能退化的现场分析", "summary_en": "A field analysis of performance degradation in coding agents across vendors.", "linkers": [] }, { "id": "50300028", "url": "https://institute.deepmind.com/essays/what-is-work-for-human-wellbeing-and-the-value-of-work-in-the-agi-era/", "domain": "institute.deepmind.com", "title": "What is work for? Human wellbeing and the value of work in the AGI era", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50300028", "points": 3, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "DeepMind研究AGI时代工作与幸福", "summary_en": "DeepMind studies work and wellbeing in the era of AGI.", "linkers": [] }, { "id": "50336359", "url": "https://www.fratepietro.com/2026/how-to-build-a-decision-model-like-jev/", "domain": "fratepietro.com", "title": "How to Build a Decision Model Like Jev", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50336359", "points": 3, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "讲解如何构建快速输出校准概率的决策模型", "summary_en": "Explains how to build a decision model that quickly outputs calibrated probabilities.", "linkers": [] }, { "id": "50342337", "url": "https://prefrontal.ai/news/lsa/", "domain": "prefrontal.ai", "title": "Accelerating Attention for Ultra-Long Context", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50342337", "points": 3, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "提出 LSA 层级聚类注意力以加速超长上下文", "summary_en": "Proposes LSA, a hierarchical clustering attention, to speed up ultra-long context.", "linkers": [] }, { "id": "50362451", "url": "https://hxu296.github.io/ai-compute-landscape/", "domain": "hxu296.github.io", "title": "Will AI compute stay tight?", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50362451", "points": 3, "comments": 1, "date": "2026-11-02", "submissions": 2, "source": "hn", "summary_zh": "分析 AI 算力租金与承诺以判断 neocloud 是否泡沫", "summary_en": "Analyzes AI compute rents and commitments to judge whether neoclouds are a bubble.", "linkers": [] }, { "id": "50348472", "url": "https://archytas-ai.com/blog/this-is-fine", "domain": "archytas-ai.com", "title": "“This Is Fine”: How Clean Rooms Train Agents to Close the Books on Fire", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50348472", "points": 3, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "研究 RL 环境如何让智能体虚报任务完成", "summary_en": "Studies how RL environments lead agents to falsely report task completion.", "linkers": [] }, { "id": "50353899", "url": "https://huggingface.co/blog/ibm-research/cuga-harness", "domain": "huggingface.co", "title": "Same Model, Different Harness: CUGA, Hermes, and Deep Agents", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50353899", "points": 3, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "比较 CUGA、Hermes、Deep Agents 三种 agent harness", "summary_en": "Compares three agent harnesses: CUGA, Hermes, and Deep Agents.", "linkers": [] }, { "id": "50354591", "url": "https://www.coreauto.com/hitchhikers-guide-to-training-on-compute-islands", "domain": "coreauto.com", "title": "The Hitchhiker’s Guide to Training on Compute Islands", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50354591", "points": 3, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "跨数据中心容错分布式训练的立场与稀疏 MoE 实验", "summary_en": "A position on fault-tolerant distributed training across data centers, with sparse MoE experiments.", "linkers": [] }, { "id": "50371607", "url": "https://devblogs.microsoft.com/foundry-on-windows/coding-ai-on-windows-pcs-quality-efficiency/", "domain": "devblogs.microsoft.com", "title": "Advancing the Quality-Efficiency Frontier for Coding AI on Windows PCs - Microsoft Foundry on Windows Blog Coding AI on Windows: The Quality–Efficiency Frontier", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50371607", "points": 3, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "用低比特量化与推测解码提升 Windows 本地编码 AI 效率", "summary_en": "Uses low-bit quantization and speculative decoding to improve local coding AI efficiency on Windows.", "linkers": [] }, { "id": "50372064", "url": "https://runanywhere.ai/blog/glm-5-3-flash-trainium2-nki-megakernels", "domain": "runanywhere.ai", "title": "Serving GLM-5.3-Flash on one Trainium2 box with NKI layer megakernels", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50372064", "points": 3, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "在单台 Trainium2 上服务 GLM-5.3-Flash 并测量吞吐", "summary_en": "Serves GLM-5.3-Flash on a single Trainium2 machine and measures throughput.", "linkers": [] }, { "id": "50375757", "url": "https://wenhaochai.com/blogs/vocab-multi-epoch.html", "domain": "wenhaochai.com", "title": "Larger Tokenizers Suffer More Multi-Epoch Regret", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50375757", "points": 3, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "实验显示更大的分词器多轮训练遗憾更高", "summary_en": "Experiments show that larger tokenizers incur higher regret in multi-epoch training.", "linkers": [] }, { "id": "50030271", "url": "https://veds12.github.io/git-swarm/", "domain": "veds12.github.io", "title": "GitSwarm: Decentralized Compounding Inference", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50030271", "points": 2, "comments": 0, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "提出GitSwarm，让代理群共享Git仓库累积尝试", "summary_en": "Proposes GitSwarm, in which a swarm of agents shares a Git repository to accumulate attempts.", "linkers": [] }, { "id": "50052902", "url": "https://nvlabs.github.io/Sana/Sol-Engine/StreamRender-H3/", "domain": "nvlabs.github.io", "title": "StreamRender-H3: Interactive Coding Worlds with Real-time Streaming Diffusion Render", "category": "research_report", "topic": "robotics-graphics", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50052902", "points": 2, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "提出StreamRender-H3，用流式扩散渲染交互代码世界", "summary_en": "Proposes StreamRender-H3, which renders interactive coding worlds with streaming diffusion.", "linkers": [ "huggingface.co" ] }, { "id": "50076599", "url": "https://sunovivid.github.io/correction-training/", "domain": "sunovivid.github.io", "title": "Correction Training for Diffusion Models", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50076599", "points": 2, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "提出扩散模型的 correction training 以缓解曝光偏差", "summary_en": "Proposes correction training for diffusion models to mitigate exposure bias.", "linkers": [] }, { "id": "50125263", "url": "https://igort99.github.io/posts/2026/10/05/looped-cot-faithfulness/", "domain": "igort99.github.io", "title": "Do looped transformers write reasoning they don't use?", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50125263", "points": 2, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "实验检验 looped transformer Huginn 是否真正使用其推理", "summary_en": "Experiments test whether the looped transformer Huginn actually uses its reasoning.", "linkers": [] }, { "id": "50171433", "url": "https://nickdrozd.github.io/2026/10/20/blanking-beavers-results-and-open-problems.html", "domain": "nickdrozd.github.io", "title": "Blanking Beavers: Results and Open Problems", "category": "research_report", "topic": "math-theory", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50171433", "points": 2, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "报告 Blanking Beavers 图灵机问题的结果与开放问题", "summary_en": "Reports results and open problems for the Blanking Beavers Turing machine problem.", "linkers": [ "bbchallenge.org" ] }, { "id": "50182533", "url": "https://jeff-kline.github.io/posts/economics-nobel-2026/index.html", "domain": "jeff-kline.github.io", "title": "Stabilizing the economy through expectations", "category": "explainer", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50182533", "points": 2, "comments": 0, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "交互式模型讲解 2026 年诺贝尔经济学奖的预期与衰退", "summary_en": "An interactive model explaining expectations and recessions in the 2026 Nobel Prize in Economics.", "linkers": [] }, { "id": "50199236", "url": "https://heyyviv.github.io/blog/what-reel-you-will-get-next/", "domain": "heyyviv.github.io", "title": "What Reel You Will Get Next", "category": "explainer", "topic": "ai-ml", "confidence": "low", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50199236", "points": 2, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "通过观察网络请求讲解短视频推荐系统设计", "summary_en": "Explains short-video recommendation system design by observing network requests.", "linkers": [] }, { "id": "50212282", "url": "https://zzyzek.github.io/SolidGridGraphHamPathPaper/", "domain": "zzyzek.github.io", "title": "Hamiltonian Paths in Solid Grid Graphs", "category": "research_report", "topic": "math-theory", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50212282", "points": 2, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "证明实心网格图哈密顿路径可多项式时间判定", "summary_en": "Proves that Hamiltonian paths in solid grid graphs can be decided in polynomial time.", "linkers": [] }, { "id": "50221659", "url": "https://sandyresearch.github.io/benice/", "domain": "sandyresearch.github.io", "title": "BeNice! Closing the Inference-Memory Gap of Looped Models with KV-Sharing–Sandy Research", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50221659", "points": 2, "comments": 0, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "通过KV-sharing缩小循环模型推理内存差距的研究", "summary_en": "Research on narrowing the inference-memory gap of looped models through KV-sharing.", "linkers": [] }, { "id": "50278937", "url": "https://daveremy.github.io/agent-scrum/", "domain": "daveremy.github.io", "title": "Agent Scrum", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50278937", "points": 2, "comments": 2, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "AI编码智能体冲刺方法及324次冲刺观察", "summary_en": "An agent sprint method for AI coding agents, with observations from 324 sprints.", "linkers": [] }, { "id": "50278943", "url": "https://yifanzhang-pro.github.io/recurrent-looped-tranformer/", "domain": "yifanzhang-pro.github.io", "title": "Recurrent Looped Transformer", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": true, "hn_id": "50278943", "points": 2, "comments": 1, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "Recurrent Looped Transformer状态跟踪项目页", "summary_en": "Project page on state tracking with the Recurrent Looped Transformer.", "linkers": [] }, { "id": "50293196", "url": "https://fsdatalab.github.io/blog/agent-text-at-scale/", "domain": "fsdatalab.github.io", "title": "Five Principles for Generating Readable AI Text in Bulk", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50293196", "points": 2, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "批量生成可读AI文本的五条原则", "summary_en": "Five principles for generating readable AI text in bulk.", "linkers": [] }, { "id": "50295254", "url": "https://jax-ml.github.io/scaling-book/", "domain": "jax-ml.github.io", "title": "How To Scale Your Model", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50295254", "points": 2, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "介绍TPU上LLM扩展的系统视角教程", "summary_en": "A systems-perspective tutorial on scaling LLMs on TPUs.", "linkers": [] }, { "id": "50332213", "url": "https://tencent.github.io/WeVisCut", "domain": "tencent.github.io", "title": "WeVisCut: The Whole Editing Room in One Agent", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50332213", "points": 2, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "WeVisCut 项目页，用智能体完成长视频剪辑", "summary_en": "Project page for WeVisCut, which uses an agent to edit long videos.", "linkers": [] }, { "id": "50344737", "url": "https://maknee.github.io/blog/2026/Cuda-Checkpointing/", "domain": "maknee.github.io", "title": "Making CUDA Checkpointing 15x Faster on a B300: A Worklog", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50344737", "points": 2, "comments": 0, "date": "2026-11-02", "submissions": 2, "source": "hn", "summary_zh": "记录将 B300 上 CUDA 检查点加速 15 倍的工作日志", "summary_en": "A worklog on making CUDA checkpointing 15 times faster on a B300.", "linkers": [] }, { "id": "50350119", "url": "https://0x4d31.github.io/airt/", "domain": "0x4d31.github.io", "title": "AIRT—AI Red Team Academy", "category": "explainer", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50350119", "points": 2, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "免费开源的 AI 系统红队攻防课程", "summary_en": "A free, open-source course on red teaming and defending AI systems.", "linkers": [] }, { "id": "50362983", "url": "https://timvieira.github.io/testing-randomized-algorithms/", "domain": "timvieira.github.io", "title": "Ideas for Testing Randomized Algorithms", "category": "explainer", "topic": "math-theory", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50362983", "points": 2, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "介绍随机算法的测试思路", "summary_en": "Introduces ideas for testing randomized algorithms.", "linkers": [] }, { "id": "50378788", "url": "https://anirudhb11.github.io/blog/scorpio/", "domain": "anirudhb11.github.io", "title": "On the Limits of Test-Time Compute for Self-Evaluation", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50378788", "points": 2, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "分析测试时计算用于自我评估的局限", "summary_en": "Analyzes the limits of using test-time compute for self-evaluation.", "linkers": [] }, { "id": "50380700", "url": "https://jonneslin.github.io/weight-space-trigger-en.html", "domain": "jonneslin.github.io", "title": "Hiding a Rule Inside Another Model's Weights", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50380700", "points": 2, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "手动编辑 Qwen3.5-0.8B 约 0.1% 权重植入规则", "summary_en": "Manually edits about 0.1% of Qwen3.5-0.8B's weights to implant a rule.", "linkers": [] }, { "id": "49998469", "url": "https://imhgchoi.github.io/agentic-idea-manager/", "domain": "imhgchoi.github.io", "title": "AIM — Agentic Idea Management for Automated Research", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "49998469", "points": 1, "comments": 0, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "提出AIM，用代理管理和探索自动化研究的想法", "summary_en": "Proposes AIM, which uses agents to manage and explore ideas for automated research.", "linkers": [] }, { "id": "49999871", "url": "https://infini-ai-lab.github.io/ServeLearnBench/", "domain": "infini-ai-lab.github.io", "title": "ServeLearnBench", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "49999871", "points": 1, "comments": 0, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "提出ServeLearnBench，评测代理从服务经验中自我改进", "summary_en": "Proposes ServeLearnBench to evaluate agents' self-improvement from serving experience.", "linkers": [] }, { "id": "50004628", "url": "https://datascienceconfidential.github.io//economics/2026/10/04/induced-demand.html", "domain": "datascienceconfidential.github.io", "title": "Some simple models of induced demand", "category": "explainer", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50004628", "points": 1, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "用简单模型解释诱发需求现象", "summary_en": "Explains the phenomenon of induced demand with simple models.", "linkers": [ "substack.com" ] }, { "id": "50021580", "url": "https://efficient-agents.github.io/", "domain": "efficient-agents.github.io", "title": "Toward Efficient Agents: A Survey of Memory, Tool learning, and Planning", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": true, "hn_id": "50021580", "points": 1, "comments": 0, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "综述高效代理的记忆、工具学习与规划", "summary_en": "A survey of memory, tool learning, and planning for efficient agents.", "linkers": [] }, { "id": "50032037", "url": "https://itsodeleo.github.io/posts/local-decision-workflow/", "domain": "itsodeleo.github.io", "title": "If LLMs Can Decide Without Fine-Tuning, Do We Still Need Models Like Jev?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50032037", "points": 1, "comments": 0, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "本地测试Qwen等LLM无需微调做决策与Jev对比", "summary_en": "Locally tests LLMs such as Qwen for decision making without fine-tuning and compares them with Jev.", "linkers": [] }, { "id": "50080086", "url": "https://astralord.github.io/posts/speculative-decoding-the-art-of-guessing-well/", "domain": "astralord.github.io", "title": "Speculative Decoding: The Art of Guessing Well", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50080086", "points": 1, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "讲解 speculative decoding 如何加速 LLM 解码", "summary_en": "Explains how speculative decoding speeds up LLM decoding.", "linkers": [] }, { "id": "50099321", "url": "https://trainstation-project.github.io/blog/flat/", "domain": "trainstation-project.github.io", "title": "FLAT: Flash Linear Attention on TPUs", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50099321", "points": 1, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "在 TPU 上用 Pallas 实现 Linear Attention 等 kernel", "summary_en": "Implements kernels such as linear attention on TPUs using Pallas.", "linkers": [] }, { "id": "50099465", "url": "https://catarinagamboa.github.io/papers/barista/", "domain": "catarinagamboa.github.io", "title": "Barista: Synthesizing Typestate Specifications with LLM Agents", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50099465", "points": 1, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "Barista：用 LLM agent 合成 typestate 规约（ASE 2026）", "summary_en": "Barista: synthesizes typestate specifications with LLM agents (ASE 2026).", "linkers": [] }, { "id": "50109691", "url": "https://kellerjordan.github.io/posts/muon/", "domain": "kellerjordan.github.io", "title": "Muon: An optimizer for hidden layers in neural networks", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50109691", "points": 1, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "介绍用于隐藏层的 Muon 优化器及其训练速度记录", "summary_en": "Introduces the Muon optimizer for hidden layers and its training speed records.", "linkers": [ "deep-learning-alchemy.github.io", "lucastononro.github.io", "xingjianbai.com" ] }, { "id": "50131292", "url": "https://chessapig.github.io/Mahler", "domain": "chessapig.github.io", "title": "Illustrating the AI proof of the Mahler conjecture", "category": "explainer", "topic": "math-theory", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50131292", "points": 1, "comments": 0, "date": "2026-10-17", "submissions": 1, "source": "hn", "summary_zh": "图解 AI 证明 Mahler 猜想的思路", "summary_en": "Illustrates the ideas behind the AI proof of the Mahler conjecture.", "linkers": [] }, { "id": "50139448", "url": "https://keenableai.github.io/needle/", "domain": "keenableai.github.io", "title": "NEEDLE — search engine benchmarks", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50139448", "points": 1, "comments": 0, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "发布面向 agent 的开源实时搜索引擎基准 Needle 及结果", "summary_en": "Releases Needle, an open-source real-time search engine benchmark for agents, with results.", "linkers": [] }, { "id": "50165308", "url": "https://weights-exfil.github.io/model-weights-exfiltration/", "domain": "weights-exfil.github.io", "title": "How leaky is your inference API? Detecting model weight steganographic exfiltration", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50165308", "points": 1, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "展示通过 LLM API 的 token 选择窃取模型权重的隐蔽信道", "summary_en": "Shows a covert channel that steals model weights through token choices in LLM APIs.", "linkers": [] }, { "id": "50194415", "url": "https://baidu-baige.github.io/LoongForge/projects/vlda.html", "domain": "baidu-baige.github.io", "title": "LoongForge-VLDA", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50194415", "points": 1, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "提出 VLDA 方法，用紧凑 Delta 作为机器人操作接口", "summary_en": "Proposes VLDA, which uses compact deltas as the interface for robot manipulation.", "linkers": [] }, { "id": "50194948", "url": "https://siimtolk.github.io/articles/llm-testing-overnight/", "domain": "siimtolk.github.io", "title": "Automated nightly LLM testing: an AI agent benchmarks new local models while I sleep", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50194948", "points": 1, "comments": 1, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "在两张 RTX 5060 Ti 上测试 18 个本地 LLM", "summary_en": "Tests 18 local LLMs on two RTX 5060 Ti GPUs.", "linkers": [] }, { "id": "50212268", "url": "https://zzyzek.github.io/R2K2ZZNPaper/", "domain": "zzyzek.github.io", "title": "Two Path Zig Zag Numberlink", "category": "research_report", "topic": "math-theory", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50212268", "points": 1, "comments": 2, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "给出 Numberlink 的禁止模式刻画与多项式算法", "summary_en": "Gives a forbidden-pattern characterization of Numberlink and a polynomial-time algorithm.", "linkers": [] }, { "id": "50220510", "url": "https://leggedrobotics.github.io/rubik-skills/", "domain": "leggedrobotics.github.io", "title": "Solving a Giant Rubik's Cube with a Quadruped Robot", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50220510", "points": 1, "comments": 1, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "四足机器人自主扫描并解开巨型魔方的项目页", "summary_en": "Project page on a quadruped robot that autonomously scans and solves a giant Rubik's cube.", "linkers": [] }, { "id": "50260492", "url": "https://leesintheblindmonk1999.github.io/sas-landing/#hero", "domain": "leesintheblindmonk1999.github.io", "title": "SAS|Symbiotic Autoprotection System|κD = 0.56", "category": "research_report", "topic": "ai-ml", "confidence": "low", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50260492", "points": 1, "comments": 0, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "声称的确定性AI结构审计框架K_D=0.56", "summary_en": "A claimed deterministic AI structural audit framework with K_D=0.56.", "linkers": [] }, { "id": "50271643", "url": "https://nvlabs.github.io/motionbricks/", "domain": "nvlabs.github.io", "title": "MotionBricks: Scalable Real-Time Motions with Modular Latent Generative Model and Smart Primitives", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50271643", "points": 1, "comments": 0, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "MotionBricks实时运动生成论文项目页", "summary_en": "Project page for the MotionBricks real-time motion generation paper.", "linkers": [] }, { "id": "50273511", "url": "https://lucastononro.github.io/how-you-could-have-invented-muon-from-memory/", "domain": "lucastononro.github.io", "title": "how you could have invented muon from memory", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50273511", "points": 1, "comments": 0, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "从记忆角度推导Muon优化器的教程", "summary_en": "A tutorial deriving the Muon optimizer from the perspective of memory.", "linkers": [] }, { "id": "50289341", "url": "https://rajabi2001.github.io/area/", "domain": "rajabi2001.github.io", "title": "AReA", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50289341", "points": 1, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "AReA高分辨率扩散Transformer高效注意力项目页", "summary_en": "Project page for AReA, efficient attention for high-resolution diffusion transformers.", "linkers": [] }, { "id": "50343763", "url": "https://zerobench.github.io/", "domain": "zerobench.github.io", "title": "ZeroBench", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50343763", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "提出对多模态大模型极难的视觉基准 ZeroBench", "summary_en": "Proposes ZeroBench, a visual benchmark that is extremely hard for multimodal large models.", "linkers": [] }, { "id": "50350996", "url": "http://karpathy.github.io/2015/05/21/rnn-effectiveness/", "domain": "karpathy.github.io", "title": "The Unreasonable Effectiveness of Recurrent Neural Networks", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50350996", "points": 1, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "介绍 RNN 及字符级语言模型实验的经典文章", "summary_en": "A classic article introducing RNNs and character-level language model experiments.", "linkers": [] }, { "id": "50351125", "url": "https://andrew-helmer.github.io/sz-matrix-construction/", "domain": "andrew-helmer.github.io", "title": "A Direct Construction of SZ Sequence Matrices", "category": "research_report", "topic": "math-theory", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "hn_id": "50351125", "points": 1, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "给出 SZ 序列矩阵的直接构造方法", "summary_en": "Gives a direct construction of SZ sequence matrices.", "linkers": [] }, { "id": "50363655", "url": "https://flecart.github.io/posts/2026-11-03-i-tried-to-break-weight-programming/", "domain": "flecart.github.io", "title": "I Tried to Break Weight Programming", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": true, "hn_id": "50363655", "points": 1, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "复现权重编程实验并分析复杂度限制", "summary_en": "Replicates weight programming experiments and analyzes their complexity limits.", "linkers": [] }, { "id": "f0", "source": "fork", "url": "https://osmangult.github.io/blog/2026/pinn-myocardium/", "domain": "osmangult.github.io", "title": "PINN for passive myocardium hyperelasticity", "category": "research_report", "topic": "bio-health", "confidence": "low", "date": "2026-10-10", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "用 PINN 拟合被动心肌超弹性本构的项目说明", "summary_en": "A project description of fitting passive myocardium hyperelastic constitutive laws with a PINN.", "linkers": [] }, { "id": "f5", "source": "fork", "url": "https://floriajin.github.io/blog/2026/is-the-multi-agent-system-another-hallucination/", "domain": "floriajin.github.io", "title": "Is the Multi-Agent System Another Hallucination?", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "date": "2026-10-28", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "综述 LLM 智能体与多智能体系统的故障诊断研究", "summary_en": "A survey of failure diagnosis research for LLM agents and multi-agent systems.", "linkers": [] }, { "id": "f6", "source": "fork", "url": "https://panreta.github.io/blog/2026/KLDivergence/", "domain": "panreta.github.io", "title": "KL-Divergence and Entropy", "category": "explainer", "topic": "math-theory", "confidence": "high", "date": "2026-11-01", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "讲解 Bregman 散度、KL 散度与熵的来源和性质", "summary_en": "Explains the origins and properties of Bregman divergence, KL divergence, and entropy.", "linkers": [] }, { "id": "f7", "source": "fork", "url": "https://dongxuanhe.github.io/blog/2026/toward-dexterous-mobile-manipulation/", "domain": "dongxuanhe.github.io", "title": "Toward Dexterous Mobile Manipulation", "category": "explainer", "topic": "robotics-graphics", "confidence": "medium", "date": "2026-10-30", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "记录 Nikolaus Correll 关于灵巧移动操作的讲座", "summary_en": "Notes from Nikolaus Correll's talk on dexterous mobile manipulation.", "linkers": [] }, { "id": "f8", "source": "fork", "url": "https://dongxuanhe.github.io/blog/2026/statistical-inference-with-ai-generated-predictive-distributions/", "domain": "dongxuanhe.github.io", "title": "Statistical Inference with AI-Generated Predictive Distributions", "category": "explainer", "topic": "math-theory", "confidence": "medium", "date": "2026-10-30", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "记录 Emma Zhang 关于 AI 预测分布统计推断的讲座", "summary_en": "Notes from Emma Zhang's talk on statistical inference with AI predictive distributions.", "linkers": [] }, { "id": "f12", "source": "fork", "url": "https://dongxuanhe.github.io/blog/2026/identifying-discrete-state-network-dynamics/", "domain": "dongxuanhe.github.io", "title": "Identifying Discrete-State Network Dynamics", "category": "explainer", "topic": "math-theory", "confidence": "medium", "date": "2026-10-23", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "记录 Carolyn Beck 关于离散状态网络动态辨识的研讨会", "summary_en": "Notes from Carolyn Beck's seminar on identifying discrete-state network dynamics.", "linkers": [] }, { "id": "f13", "source": "fork", "url": "https://dongxuanhe.github.io/blog/2026/learning-causal-graphs-with-mixed-integer-optimization/", "domain": "dongxuanhe.github.io", "title": "Learning Causal Graphs with Mixed-Integer Optimization", "category": "explainer", "topic": "math-theory", "confidence": "medium", "date": "2026-10-16", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "记录 Küçükyavuz 关于混合整数优化学习因果图的研讨会", "summary_en": "Notes from Küçükyavuz's seminar on learning causal graphs with mixed-integer optimization.", "linkers": [] }, { "id": "f14", "source": "fork", "url": "https://wenyaxie023.github.io/blog/2026/ml-conference-trends/", "domain": "wenyaxie023.github.io", "title": "ML Conference Trends: NeurIPS & ICLR, 2023–2026", "category": "research_report", "topic": "ai-ml", "confidence": "high", "date": "2026-10-15", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "统计 NeurIPS 与 ICLR 2023-2026 录用论文的主题趋势", "summary_en": "Counts topic trends in accepted NeurIPS and ICLR papers from 2023 to 2026.", "linkers": [] }, { "id": "f16", "source": "fork", "url": "https://mayaokawa.github.io/blog/2026/biased-consensus/", "domain": "mayaokawa.github.io", "title": "When Do Innocent LLM Agents Reach Unfair Decisions?", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "date": "2026-10-10", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "用 Ising 模型分析 LLM 辩论中微小偏见放大为不公平共识", "summary_en": "Uses an Ising model to analyze how small biases amplify into unfair consensus in LLM debates.", "linkers": [] }, { "id": "f17", "source": "fork", "url": "https://aopatric.github.io/blog/2026/recurrent-grokking/", "domain": "aopatric.github.io", "title": "On Serial Depth and Monitorability in Looped Transformers, Part 1", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "date": "2026-10-23", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "研究循环 Transformer 的串行深度与思维链可监控性", "summary_en": "Studies serial depth in looped transformers and the monitorability of chain of thought.", "linkers": [] }, { "id": "f18", "source": "fork", "url": "https://from-sea-to-summit.github.io/blog/2026/evaporation-measurements/", "domain": "from-sea-to-summit.github.io", "title": "Evapotranspiration monitoring", "category": "research_report", "topic": "physics-earth", "confidence": "low", "date": "2026-10-09", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "记录蒸散发监测设备的安装与初步观测", "summary_en": "Records the installation of evapotranspiration monitoring equipment and initial observations.", "linkers": [] }, { "id": "f23", "source": "fork", "url": "https://bishnubhusal.com.np/blog/2026/clipping-geometry-private-muon/", "domain": "bishnubhusal.com.np", "title": "Why gradient clipping changes Muon’s geometry: a two-dimensional example", "category": "research_report", "topic": "ai-ml", "confidence": "high", "date": "2026-10-17", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "blog": null, "doi": null, "summary_zh": "用二维例子说明梯度裁剪如何改变 Muon 的几何", "summary_en": "Uses a two-dimensional example to show how gradient clipping changes Muon's geometry.", "linkers": [] }, { "id": "f24", "source": "fork", "url": "https://schiekiera.github.io/blog/2026/computational-swinging-lexical-network/", "domain": "schiekiera.github.io", "title": "A Computational Swinging Lexical Network", "category": "research_report", "topic": "humanities", "confidence": "high", "date": "2026-10-15", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "实现词语产出的 Swinging Lexical Network 计算模型", "summary_en": "Implements a computational model of the Swinging Lexical Network for word production.", "linkers": [] }, { "id": "f25", "source": "fork", "url": "https://schiekiera.github.io/blog/2026/linear-pfn/", "domain": "schiekiera.github.io", "title": "LinearPFN: Amortized Variable Selection for Linear Models with Interactions", "category": "research_report", "topic": "ai-ml", "confidence": "high", "date": "2026-10-12", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": true, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "提出预训练的 LinearPFN 用于含交互项线性模型的变量选择", "summary_en": "Proposes a pretrained LinearPFN for variable selection in linear models with interactions.", "linkers": [] }, { "id": "f26", "source": "fork", "url": "https://chandlersmith2.github.io/blog/2026/decoupling-inequality-proof-sketch/", "domain": "chandlersmith2.github.io", "title": "A Proof Sketch of the Inequality of de la Peña and Giné", "category": "explainer", "topic": "math-theory", "confidence": "high", "date": "2026-11-02", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "给出 de la Peña 不等式的证明思路", "summary_en": "Gives a proof sketch of the inequality of de la Peña.", "linkers": [] }, { "id": "f31", "source": "fork", "url": "https://boxiyu.github.io/writing/what-frontiercs-2-teaches-about-agent-results/", "domain": "boxiyu.github.io", "title": "What FrontierCS 2 teaches about reading agent results", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "date": "2026-10-28", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "讨论 FrontierCS 2 基准的通过规则对解读智能体结果的启示", "summary_en": "Discusses what the pass rules of the FrontierCS 2 benchmark imply for interpreting agent results.", "linkers": [] }, { "id": "f33", "source": "fork", "url": "https://zlatanajanovic.com/robot-autonomy-levels/", "domain": "zlatanajanovic.com", "title": "Who is in charge? Robot autonomy levels for human-in-the-loop systems", "category": "research_report", "topic": "robotics-graphics", "confidence": "medium", "date": "2026-10-28", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": null, "doi": null, "summary_zh": "提出人在回路系统的机器人自主等级框架（书章节）", "summary_en": "Proposes a framework of robot autonomy levels for human-in-the-loop systems (book chapter).", "linkers": [] }, { "id": "f44", "source": "fork", "url": "https://fpoltronieri.github.io/blog/2026/teaching-a-car-to-drive-inside-its-own-dream/", "domain": "fpoltronieri.github.io", "title": "teaching a car to drive inside its own dream", "category": "research_report", "topic": "robotics-graphics", "confidence": "medium", "date": "2026-10-27", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "记录让自动驾驶在潜在世界模型内学会驾驶的实验", "summary_en": "Records experiments teaching autonomous driving to drive inside a latent world model.", "linkers": [] }, { "id": "f46", "source": "fork", "url": "https://elena6918.github.io/blog/2026/rules/", "domain": "elena6918.github.io", "title": "When Security Rules Keep Changing Their Minds", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "date": "2026-10-17", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "研究 LLM 编写的安全检测规则在迭代中的版本反复", "summary_en": "Studies version flip-flopping in LLM-written security detection rules across iterations.", "linkers": [] }, { "id": "f53", "source": "fork", "url": "https://eureka10shen.github.io/blog/quantum-computation-quantum-chemistry/", "domain": "eureka10shen.github.io", "title": "QC^2 : Quantum Computation for Quantum Chemistry", "category": "explainer", "topic": "physics-earth", "confidence": "medium", "date": "2026-10-08", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "量子化学量子计算第一步的技术笔记", "summary_en": "Technical notes on the first step of quantum computation for quantum chemistry.", "linkers": [] }, { "id": "f55", "source": "fork", "url": "https://anwoybarua.github.io/posts/2026/10/xarray-guidebook-part1/", "domain": "anwoybarua.github.io", "title": "Xarray Guidebook Part 1: Working with Multidimensional Climate Data", "category": "explainer", "topic": "physics-earth", "confidence": "high", "date": "2026-10-07", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "用 Xarray 处理多维气候数据的入门教程", "summary_en": "An introductory tutorial on processing multidimensional climate data with Xarray.", "linkers": [] }, { "id": "f62", "source": "fork", "url": "https://daudkhan2030.github.io/engineering/2026/10/23/Multithreading_Mutex_Thread_Synchronization.html", "domain": "daudkhan2030.github.io", "title": "C++ Multithreading: Mutex, Condition Variables & Thread Synchronization", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "date": "2026-10-23", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "讲解 C++ 多线程中的互斥锁与条件变量", "summary_en": "Explains mutexes and condition variables in C++ multithreading.", "linkers": [] }, { "id": "f76", "source": "fork", "url": "https://inamullah-colab.github.io/posts/2026/10/risk-odds-and-causality/", "domain": "inamullah-colab.github.io", "title": "Risk, Odds, and Causality: A Practical Guide for Epidemiology and Healthcare AI", "category": "explainer", "topic": "bio-health", "confidence": "high", "date": "2026-10-25", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "讲解流行病学与医疗 AI 中的风险、比值比与因果", "summary_en": "Explains risk, odds ratios, and causality in epidemiology and healthcare AI.", "linkers": [] }, { "id": "f77", "source": "fork", "url": "https://ravulapallimanas.github.io/writing/2026/10/ideas-on-creating-artificial-life-in-silico/", "domain": "ravulapallimanas.github.io", "title": "Lessons from building in silico life with homeostatic RL", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "date": "2026-10-08", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "总结用稳态强化学习构建体内需求型人工生命的经验", "summary_en": "Summarizes lessons from building need-driven artificial life with homeostatic reinforcement learning.", "linkers": [] }, { "id": "f80", "source": "fork", "url": "https://markssce.github.io/Mercury.github.io/posts/2026/10/blog-post-7/", "domain": "markssce.github.io", "title": "20261028", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "date": "2026-10-28", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "LSTM 与深度学习的学习笔记", "summary_en": "Study notes on LSTMs and deep learning.", "linkers": [] }, { "id": "f85", "source": "fork", "url": "https://markssce.github.io/Mercury.github.io/posts/2026/10/blog-post-2/", "domain": "markssce.github.io", "title": "20261021", "category": "explainer", "topic": "humanities", "confidence": "low", "date": "2026-10-21", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "说明学生先验知识对学习的促进与阻碍", "summary_en": "Explains how students' prior knowledge can both help and hinder learning.", "linkers": [] }, { "id": "f87", "source": "fork", "url": "https://muzi2018.github.io/supermimic/", "domain": "muzi2018.github.io", "title": "读 SuperMimic：用可微世界模型做人形全身跟踪，闭环是怎么转起来的", "category": "explainer", "topic": "robotics-graphics", "confidence": "medium", "date": "2026-10-22", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "解读 SuperMimic 用可微世界模型做人形全身跟踪", "summary_en": "Explains how SuperMimic uses a differentiable world model for humanoid whole-body tracking.", "linkers": [] }, { "id": "f88", "source": "fork", "url": "https://muzi2018.github.io/system-0-1-2/", "domain": "muzi2018.github.io", "title": "从 System 0 / 1 / 2 看机器人控制", "category": "explainer", "topic": "robotics-graphics", "confidence": "medium", "date": "2026-10-17", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "用 System 0/1/2 框架解释机器人控制分工", "summary_en": "Uses the System 0/1/2 framework to explain the division of labor in robot control.", "linkers": [] }, { "id": "f91", "source": "fork", "url": "https://nunobrazz.github.io/posts/2026/canton-architecture/", "domain": "nunobrazz.github.io", "title": "Canton architecture", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "date": "2026-10-31", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "介绍Canton架构中后端、参与节点与同步器如何协作", "summary_en": "Introduces how the backend, participant nodes, and synchronizers cooperate in the Canton architecture.", "linkers": [] }, { "id": "f92", "source": "fork", "url": "https://nunobrazz.github.io/posts/2026/canton-eutxo-vs-account-model/", "domain": "nunobrazz.github.io", "title": "eUTXO vs account model in Canton", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "date": "2026-10-31", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "对比以太坊账户模型与Canton的eUTXO模型", "summary_en": "Compares the Ethereum account model with Canton's eUTXO model.", "linkers": [] }, { "id": "f93", "source": "fork", "url": "https://nunobrazz.github.io/posts/2026/canton-decentralization/", "domain": "nunobrazz.github.io", "title": "How Canton decentralizes", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "date": "2026-10-31", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "说明Canton在协议层与应用层的去中心化方式", "summary_en": "Explains how Canton decentralizes at the protocol layer and the application layer.", "linkers": [] }, { "id": "f95", "source": "fork", "url": "https://geo-huanwang.com/posts/2026/10/monopile-moment-rotation-predictor/", "domain": "geo-huanwang.com", "title": "A Generalized Scaling Law for Monopiles in Sand: From Centrifuge Tests to an Interactive Predictor", "category": "research_report", "topic": "engineering-env", "confidence": "high", "date": "2026-10-31", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": null, "doi": null, "summary_zh": "基于离心机试验提出砂土中单桩的通用缩放律并附交互预测工具", "summary_en": "Proposes a general scaling law for monopiles in sand based on centrifuge tests, with an interactive prediction tool.", "linkers": [] }, { "id": "f96", "source": "fork", "url": "https://ethan-jz.github.io//ppo-robot-reaching/", "domain": "ethan-jz.github.io", "title": "Robotic arm reaching task with reinforcement learning (PPO)", "category": "research_report", "topic": "robotics-graphics", "confidence": "medium", "date": "2026-10-20", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": null, "doi": null, "summary_zh": "用PPO强化学习部署机械臂到达任务", "summary_en": "Deploys a robotic arm reaching task using PPO reinforcement learning.", "linkers": [] }, { "id": "r3", "source": "rs", "url": "https://tritonstation.com/2026/11/05/why-the-search-for-particle-dark-matter-is-really-hopeless-part-2/", "domain": "tritonstation.com", "title": "Why the search for particle dark matter is really hopeless, part 2", "category": "research_report", "topic": "physics-earth", "confidence": "medium", "date": "2026-11-05", "points": 2, "comments": 0, "hn_id": "50385821", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "blog": "Triton Station", "doi": "10.59350/kfs0g-sm298", "summary_zh": "论证粒子暗物质搜寻在MOND现象下希望渺茫（第二部分）", "summary_en": "Argues that the search for particle dark matter is nearly hopeless given MOND phenomena (part 2).", "linkers": [] }, { "id": "r19", "source": "rs", "url": "https://www.leidenmadtrics.nl/articles/tracking-r-d-funding-in-the-global-south-challenges-and-opportunities-in-india", "domain": "leidenmadtrics.nl", "title": "Tracking R&D funding in the Global South: challenges and opportunities in India", "category": "research_report", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-11-02", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Leiden Madtrics", "doi": "10.59350/w0b63-9mg44", "summary_zh": "分析印度研发经费追踪的挑战与机会", "summary_en": "Analyzes the challenges and opportunities of tracking R&D funding in India.", "linkers": [] }, { "id": "r23", "source": "rs", "url": "https://chem-bla-ics.linkedchemistry.info/2026/11/01/creating-preferred-iupac-names.html", "domain": "chem-bla-ics.linkedchemistry.info", "title": "Creating Preferred IUPAC names", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "date": "2026-11-01", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "chem-bla-ics", "doi": "10.59350/v7f5q-e2t67", "summary_zh": "探讨用语言模型生成IUPAC优选名称", "summary_en": "Explores using language models to generate preferred IUPAC names.", "linkers": [] }, { "id": "r25", "source": "rs", "url": "https://jonswain.github.io/morg2smiles/", "domain": "jonswain.github.io", "title": "Morg2SMILES: Generative Recovery of SMILES Strings from Morgan Fingerprints", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "date": "2026-11-01", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": true, "cit": false, "blog": "Jon Swain", "doi": "10.59350/srwee-de037", "summary_zh": "讨论从Morgan指纹生成式恢复SMILES字符串", "summary_en": "Discusses generative recovery of SMILES strings from Morgan fingerprints.", "linkers": [] }, { "id": "r27", "source": "rs", "url": "https://grantwitness.org/cdc/updates/methodology", "domain": "grantwitness.org", "title": "CDC Grant Disruption Methodology", "category": "research_report", "topic": "econ-social", "confidence": "high", "date": "2026-10-30", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Grant Witness", "doi": "10.59350/kscq9-c2n75", "summary_zh": "说明CDC资助中断数据的来源与方法", "summary_en": "Explains the sources and methods of the CDC grant disruption data.", "linkers": [] }, { "id": "r36", "source": "rs", "url": "https://irights.info/artikel/its-all-in-the-mix-labeling-and-licensing-materials-with-ai-content/32937", "domain": "irights.info", "title": "It's All in the Mix: Labeling and Licensing Materials with AI Content", "category": "explainer", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-29", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "iRights.info", "doi": "10.59350/4r37a-5q603", "summary_zh": "说明含AI内容的开放教育资源如何标注与授权", "summary_en": "Explains how to label and license open educational resources that contain AI content.", "linkers": [] }, { "id": "r40", "source": "rs", "url": "https://blog.tib.eu/2026/10/29/faq-wie-publiziere-ich-rechtssicher-fuer-die-allgemeinheit/", "domain": "blog.tib.eu", "title": "FAQ: Wie publiziere ich rechtssicher für die Allgemeinheit?", "category": "explainer", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-29", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "TIB-Blog", "doi": "10.65527/1m571-m6a83", "summary_zh": "解答如何合法地向公众发表成果", "summary_en": "Answers how to publish results for the public in a legally safe way.", "linkers": [] }, { "id": "r46", "source": "rs", "url": "https://www.scholcommlab.ca/2026/10/28/scrollytelling/", "domain": "scholcommlab.ca", "title": "Scrollytelling: What happens after you share your data?", "category": "explainer", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-28", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Scholarly Communications Lab | ScholCommLab", "doi": "10.59350/88k6w-z7p81", "summary_zh": "用scrollytelling展示数据引用与复用故事", "summary_en": "Uses scrollytelling to show the story of data citation and reuse.", "linkers": [] }, { "id": "r56", "source": "rs", "url": "https://thebibliomagician.wordpress.com/2026/10/26/ssrn-rankings-are-gone-how-to-read-your-own-numbers-without-rebuilding-them/", "domain": "thebibliomagician.wordpress.com", "title": "SSRN Rankings Are Gone. How to Read Your Own Numbers Without Rebuilding Them", "category": "explainer", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-26", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "The Bibliomagician", "doi": "10.59350/mgsjw-t8285", "summary_zh": "说明SSRN排名取消后如何解读下载数据", "summary_en": "Explains how to interpret download data after SSRN rankings were discontinued.", "linkers": [] }, { "id": "r59", "source": "rs", "url": "https://tritonstation.com/2026/10/26/why-the-search-for-particle-dark-matter-is-hopeless-part-1/", "domain": "tritonstation.com", "title": "Why the search for particle dark matter is hopeless, part 1", "category": "research_report", "topic": "physics-earth", "confidence": "medium", "date": "2026-10-26", "points": 2, "comments": 1, "hn_id": "50249203", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "blog": "Triton Station", "doi": "10.59350/cgn4p-jwf32", "summary_zh": "论证粒子暗物质搜寻为何无望（第一部分）", "summary_en": "Argues why the search for particle dark matter is hopeless (part 1).", "linkers": [] }, { "id": "r65", "source": "rs", "url": "https://blog.stephenturner.us/p/ai-peer-reviews-beat-humans", "domain": "blog.stephenturner.us", "title": "AI Reviews Might Beat Human Peer Reviews", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "date": "2026-10-24", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Paired Ends", "doi": "10.59350/006kh-wdm07", "summary_zh": "介绍AI评审可能不亚于人工同行评审的演示", "summary_en": "Introduces a demonstration that AI reviews may be no worse than human peer reviews.", "linkers": [] }, { "id": "r66", "source": "rs", "url": "https://grantwitness.org/epa/updates/methodology", "domain": "grantwitness.org", "title": "EPA Grant Disruption Methodology", "category": "research_report", "topic": "econ-social", "confidence": "high", "date": "2026-10-23", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Grant Witness", "doi": "10.59350/tq3c6-7x998", "summary_zh": "说明EPA资助中断数据的来源与方法", "summary_en": "Explains the sources and methods of the EPA grant disruption data.", "linkers": [] }, { "id": "r75", "source": "rs", "url": "https://medical-education.pages.ub.uni-bielefeld.de/research/mes-blog/posts/2026-10-22-forschendes-lernen/", "domain": "medical-education.pages.ub.uni-bielefeld.de", "title": "Forschendes Lernen", "category": "explainer", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-21", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Entscheiden(d) lernen", "doi": "10.59350/48bhv-zby17", "summary_zh": "讨论研究型学习需要引导才有效", "summary_en": "Discusses why research-based learning needs guidance to be effective.", "linkers": [] }, { "id": "r77", "source": "rs", "url": "https://www.jeremykun.com/2026/10/21/ckks-key-switching-and-relinearization/", "domain": "jeremykun.com", "title": "CKKS — Key Switching and Relinearization", "category": "explainer", "topic": "security-privacy", "confidence": "high", "date": "2026-10-21", "points": 2, "comments": 0, "hn_id": "50197247", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Math ∩ Programming", "doi": "10.59350/ge6c9-9p826", "summary_zh": "CKKS同态加密教程：密钥切换与重线性化", "summary_en": "A CKKS homomorphic encryption tutorial on key switching and relinearization.", "linkers": [] }, { "id": "r82", "source": "rs", "url": "https://fluchtforschung.net/is-migration-choice-possible-during-the-migration-course-understanding-biographical-migration-projects/", "domain": "fluchtforschung.net", "title": "Is 'migration choice' possible during the migration course? Understanding biographical migration projects", "category": "research_report", "topic": "econ-social", "confidence": "medium", "date": "2026-10-21", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Netzwerk Fluchtforschung", "doi": "10.59350/2xqa7-25k66", "summary_zh": "区分传记式迁移项目与迁移意愿并分析", "summary_en": "Distinguishes biographical migration projects from migration intentions and analyzes them.", "linkers": [] }, { "id": "r95", "source": "rs", "url": "https://blog.front-matter.de/posts/what-content-categories-are-used-in-science-blogs/", "domain": "blog.front-matter.de", "title": "What content categories are used in science blogs?", "category": "research_report", "topic": "scholarly-communication", "confidence": "high", "date": "2026-10-19", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Front Matter", "doi": "10.53731/zvzh7-e8v51", "summary_zh": "分析Rogue Scholar科学博客的内容类别并提出分类法", "summary_en": "Analyzes the content categories of Rogue Scholar science blogs and proposes a taxonomy.", "linkers": [ "svpow.com" ] }, { "id": "r100", "source": "rs", "url": "https://shkspr.mobi/blog/2026/10/una-watch-totp-mfa-on-your-wrist/", "domain": "shkspr.mobi", "title": "Una Watch TOTP - MFA on your wrist!", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "date": "2026-10-18", "points": 3, "comments": 0, "hn_id": "50140430", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Terence Eden's Blog", "doi": "10.59350/0w6m9-ym602", "summary_zh": "制作手腕上的TOTP双因素验证设备", "summary_en": "Builds a TOTP two-factor authentication device for the wrist.", "linkers": [] }, { "id": "r103", "source": "rs", "url": "https://tzovar.as/codeberg-git-pages-with-lfs/", "domain": "tzovar.as", "title": "Using Codeberg's git-pages deployment with git lfs", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "date": "2026-10-17", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Bastian Greshake Tzovaras |", "doi": "10.59350/fp56a-wym34", "summary_zh": "说明如何在Codeberg git-pages中配合git lfs部署", "summary_en": "Explains how to deploy with git lfs in Codeberg git-pages.", "linkers": [] }, { "id": "r106", "source": "rs", "url": "https://chem-bla-ics.linkedchemistry.info/2026/10/17/peer-reviewed-articles-about-biological-processes.html", "domain": "chem-bla-ics.linkedchemistry.info", "title": "Peer reviewed articles about biological processes in WikiPathways", "category": "research_report", "topic": "bio-health", "confidence": "medium", "date": "2026-10-17", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "chem-bla-ics", "doi": "10.59350/2qh9a-37895", "summary_zh": "梳理WikiPathways中关于生物过程的同行评审文章", "summary_en": "Surveys peer-reviewed articles about biological processes in WikiPathways.", "linkers": [] }, { "id": "r108", "source": "rs", "url": "https://grantwitness.org/nsf/updates/methodology", "domain": "grantwitness.org", "title": "NSF Grant Disruption Methodology", "category": "research_report", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-16", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Grant Witness", "doi": "10.59350/9zz76-xcn42", "summary_zh": "说明NSF资助终止数据的追踪方法", "summary_en": "Explains how NSF grant termination data is tracked.", "linkers": [] }, { "id": "r117", "source": "rs", "url": "https://blog.bioconductor.org/posts/2026-10-15-cytometry-in-r/", "domain": "blog.bioconductor.org", "title": "Cytometry in R and opportunities to contribute to Bioconductor", "category": "explainer", "topic": "bio-health", "confidence": "medium", "date": "2026-10-15", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Bioconductor community blog", "doi": "10.59350/ktzpn-bav35", "summary_zh": "介绍R与Bioconductor中的流式细胞术工具", "summary_en": "Introduces cytometry tools in R and Bioconductor.", "linkers": [] }, { "id": "r118", "source": "rs", "url": "https://silsesquioxane.blogspot.com/2026/10/t12-dodecameric-silsesquioxanes.html", "domain": "silsesquioxane.blogspot.com", "title": "T12 POSS Cages: D6h and D2d Dodecasilsesquioxane Isomers", "category": "research_report", "topic": "physics-earth", "confidence": "medium", "date": "2026-10-15", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Polyhedral Oligomeric Silsesquioxane (POSS) Chemistry", "doi": "10.59350/2g7cs-a3r07", "summary_zh": "讨论T12 POSS笼的D6h与D2d异构体", "summary_en": "Discusses the D6h and D2d isomers of the T12 POSS cage.", "linkers": [] }, { "id": "r125", "source": "rs", "url": "https://metadatagamechangers.com/blog/2026/10/10/visualizing-crossref-repository-histories", "domain": "metadatagamechangers.com", "title": "Visualizing Crossref Repository Histories", "category": "research_report", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-14", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Blog - Metadata Game Changers", "doi": "10.59350/weqgh-xja59", "summary_zh": "可视化Crossref仓储元数据的历史", "summary_en": "Visualizes the history of Crossref repository metadata.", "linkers": [] }, { "id": "r129", "source": "rs", "url": "https://blog.tib.eu/2026/10/14/ganz-ohne-turnschuhe-administration-mit-opsi/", "domain": "blog.tib.eu", "title": "Ganz ohne Turnschuhe: Administration mit OPSI", "category": "explainer", "topic": "systems-hardware", "confidence": "medium", "date": "2026-10-14", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "TIB-Blog", "doi": "10.65527/x5cv3-p7c13", "summary_zh": "介绍用OPSI进行IT管理", "summary_en": "Introduces IT administration with OPSI.", "linkers": [] }, { "id": "r136", "source": "rs", "url": "https://winoda.de/2026/10/13/markdown-3-mein-obsidian-setup/", "domain": "winoda.de", "title": "Markdown 3 – Mein Obsidian-Setup", "category": "explainer", "topic": "other", "confidence": "high", "date": "2026-10-13", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "WiNoDa Knowledge Lab", "doi": "10.59350/rsne9-z6161", "summary_zh": "介绍作者的Obsidian设置", "summary_en": "Introduces the author's Obsidian setup.", "linkers": [] }, { "id": "r137", "source": "rs", "url": "https://blog.pensoft.net/2026/10/13/not-only-boring-issues-what-fossil-bones-reveal-about-cretaceous-hateg-island/", "domain": "blog.pensoft.net", "title": "Not Only Boring Issues: What Fossil Bones Reveal About cretaceous Haţeg Island", "category": "research_report", "topic": "physics-earth", "confidence": "medium", "date": "2026-10-13", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Pensoft Blog", "doi": "10.59350/dbp35-yrt13", "summary_zh": "介绍化石骨骼痕迹揭示白垩纪Haţeg岛的研究", "summary_en": "Introduces research in which fossil bone traces reveal the Cretaceous Haţeg Island.", "linkers": [] }, { "id": "r138", "source": "rs", "url": "https://solomoneshun.com/posts/obsidian/", "domain": "solomoneshun.com", "title": "How I Organize What I Read, Learn and Write in Obsidian", "category": "explainer", "topic": "other", "confidence": "high", "date": "2026-10-13", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Solomon Eshun", "doi": "10.59350/3hs8z-cez14", "summary_zh": "介绍用Obsidian整理阅读与写作笔记", "summary_en": "Introduces organizing reading and writing notes with Obsidian.", "linkers": [] }, { "id": "r145", "source": "rs", "url": "https://shkspr.mobi/blog/2026/10/i-paid-people-to-try-and-follow-my-readme/", "domain": "shkspr.mobi", "title": "I paid people to try and follow my README", "category": "research_report", "topic": "other", "confidence": "medium", "date": "2026-10-11", "points": 431, "comments": 206, "hn_id": "50042219", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Terence Eden's Blog", "doi": "10.59350/41s5v-7q247", "summary_zh": "付费请人按README操作以测试文档质量", "summary_en": "Pays people to follow a README to test the quality of the documentation.", "linkers": [ "substack.com" ] }, { "id": "r147", "source": "rs", "url": "https://www.jeremykun.com/2026/10/11/ckks-addition-and-multiplication/", "domain": "jeremykun.com", "title": "CKKS — Addition and Multiplication", "category": "explainer", "topic": "security-privacy", "confidence": "high", "date": "2026-10-11", "points": 10, "comments": 0, "hn_id": "50119819", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Math ∩ Programming", "doi": "10.59350/0kwhe-aqt70", "summary_zh": "讲解CKKS同态加密的加法与乘法", "summary_en": "Explains addition and multiplication in CKKS homomorphic encryption.", "linkers": [] }, { "id": "r151", "source": "rs", "url": "https://aarontay.substack.com/p/when-ai-search-makes-up-the-query", "domain": "aarontay.substack.com", "title": "When AI Search Makes Up the Query Input, and Why That Isn't Necessarily a Problem", "category": "explainer", "topic": "ai-ml", "confidence": "high", "date": "2026-10-10", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "blog": "Aaron Tay's Musings about Librarianship", "doi": "10.59350/006gy-ah542", "summary_zh": "解释HyDE与Query2doc等AI检索查询改写", "summary_en": "Explains AI search query rewriting such as HyDE and Query2doc.", "linkers": [] }, { "id": "r152", "source": "rs", "url": "https://renebekkers.wordpress.com/2026/10/10/fast-but-incomplete-and-inaccurate-twelve-ai-powered-tools-for-literature-search/", "domain": "renebekkers.wordpress.com", "title": "Fast, but incomplete and inaccurate: twelve AI-powered tools for literature search", "category": "research_report", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-10", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Rene Bekkers", "doi": "10.59350/gvwch-yhj37", "summary_zh": "评估十二个AI文献检索工具并指出其不足", "summary_en": "Evaluates twelve AI literature search tools and points out their shortcomings.", "linkers": [] }, { "id": "r155", "source": "rs", "url": "https://grantwitness.org/nih/updates/methodology", "domain": "grantwitness.org", "title": "NIH and AHRQ Grant Disruption Methodology", "category": "research_report", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-09", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Grant Witness", "doi": "10.59350/6z1w2-jqz59", "summary_zh": "说明NIH与AHRQ资助中断数据的追踪方法", "summary_en": "Explains how NIH and AHRQ grant disruption data is tracked.", "linkers": [] }, { "id": "r158", "source": "rs", "url": "https://blog.tib.eu/2026/10/09/offen-forschen-gezielt-schuetzen/", "domain": "blog.tib.eu", "title": "Offen forschen, gezielt schützen", "category": "explainer", "topic": "scholarly-communication", "confidence": "medium", "date": "2026-10-09", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "TIB-Blog", "doi": "10.65527/hbjbb-28825", "summary_zh": "说明开放科学与知识产权保护如何兼顾", "summary_en": "Explains how open science can be reconciled with intellectual property protection.", "linkers": [] }, { "id": "r160", "source": "rs", "url": "https://x-dev.pages.jsc.fz-juelich.de/2026/10/08/scicoco-posters.html", "domain": "x-dev.pages.jsc.fz-juelich.de", "title": "Posters at SciCoCo.nrw: Training LLMs and an Integrated Approach for Benchmarking", "category": "research_report", "topic": "ai-ml", "confidence": "low", "date": "2026-10-08", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "blog": "JSC Accelerating Devices Lab", "doi": "10.34732/xdvblg-0evzb5", "summary_zh": "展示SciCoCo.nrw上LLM训练与基准测试海报", "summary_en": "Presents posters from SciCoCo.nrw on LLM training and benchmarking.", "linkers": [] }, { "id": "r161", "source": "rs", "url": "https://www.jeremykun.com/shortform/2026-10-08-1120/", "domain": "jeremykun.com", "title": "Hotword detection in FHE on GPU in 350ms", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "date": "2026-10-08", "points": 1, "comments": 0, "hn_id": "50104349", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Math ∩ Programming", "doi": "10.59350/ft4wb-1vd93", "summary_zh": "报道HEIR编译器在GPU上350ms完成FHE热词检测", "summary_en": "Reports that the HEIR compiler completes FHE hotword detection on a GPU in 350 ms.", "linkers": [] }, { "id": "r167", "source": "rs", "url": "https://medical-education.pages.ub.uni-bielefeld.de/research/mes-blog/posts/2026-10-08-ernaehrung-nuechtern/", "domain": "medical-education.pages.ub.uni-bielefeld.de", "title": "Nüchtern in den Präpkurs – schadet das dem Lernen?", "category": "research_report", "topic": "bio-health", "confidence": "low", "date": "2026-10-07", "points": 0, "comments": 0, "hn_id": null, "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "blog": "Entscheiden(d) lernen", "doi": "10.59350/nqesv-f8h43", "summary_zh": "探讨空腹进入解剖课是否影响学习", "summary_en": "Discusses whether going to the anatomy course on an empty stomach affects learning.", "linkers": [] }, { "id": "49993724", "url": "https://hazyresearch.stanford.edu/blog/2026-07-22-mlps-are-hebbians", "domain": "hazyresearch.stanford.edu", "title": "🧠 MLPs are Hebbian Memories: A Simple Recipe for Fact-Storing Transformers", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "49993724", "points": 2, "comments": 0, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "提出MLP作为Hebbian记忆存储事实的闭式构造与容量保证", "summary_en": "Presents a kernel-memory view of fact-storing MLPs with a closed-form construction and capacity guarantees.", "origin": "backfill", "linkers": [] }, { "id": "49994153", "url": "https://ngerakines.leaflet.pub/3mfceipobzk2u", "domain": "ngerakines.leaflet.pub", "title": "CIDs: What You Need to Know and Why, Part 1", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "49994153", "points": 2, "comments": 0, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "介绍Content Identifier(CID)的原理与用途", "summary_en": "Explains what content identifiers are and why they matter for content addressing.", "origin": "backfill", "linkers": [] }, { "id": "49998537", "url": "https://michaeljohncooper.com/writing/change-of-variables", "domain": "michaeljohncooper.com", "title": "Visualizing the Change-of-Variables Rule in Probability • Michael Cooper", "category": "explainer", "topic": "math-theory", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "49998537", "points": 1, "comments": 1, "date": "2026-10-07", "submissions": 1, "source": "hn", "summary_zh": "可视化讲解概率中的变量替换规则与Jacobian", "summary_en": "Interactive visual explanation of the change-of-variables rule and the Jacobian in probability.", "origin": "backfill", "linkers": [] }, { "id": "50001295", "url": "https://regularized-rsi.com/", "domain": "regularized-rsi.com", "title": "RRSI: Regularized Recursive Self-Improvement of Agent Harnesses", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50001295", "points": 1, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "提出RRSI方法，缓解自改进agent对基准的过拟合", "summary_en": "Proposes regularized recursive self-improvement to stop evolved agent harnesses overfitting their benchmark.", "origin": "backfill", "linkers": [] }, { "id": "50001866", "url": "https://alignment.openai.com/metagaming-latents/", "domain": "alignment.openai.com", "title": "Studying metagaming latents in language models", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50001866", "points": 1, "comments": 0, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "研究语言模型中与监控和奖励推理相关的metagaming潜变量", "summary_en": "Studies latent features in language models linked to reasoning about monitoring and reward.", "origin": "backfill", "linkers": [ "archytas-ai.com" ] }, { "id": "50012853", "url": "https://stefankeselj.com/entries/semi_supervision/", "domain": "stefankeselj.com", "title": "The Alchemy of Semi-Supervision", "category": "research_report", "topic": "ai-ml", "confidence": "low", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50012853", "points": 1, "comments": 1, "date": "2026-10-08", "submissions": 1, "source": "hn", "summary_zh": "研究半监督学习方法并给出实验", "summary_en": "Studies semi-supervised learning as an under-used ML method with experiments.", "origin": "backfill", "linkers": [] }, { "id": "50019040", "url": "https://arena.ai/blog/coding-agents-harness-tax", "domain": "arena.ai", "title": "HarnessTax: How Much Does the Harness Matter for Coding Agents?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50019040", "points": 1, "comments": 0, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "评测21种编程agent的模型与harness组合", "summary_en": "Benchmarks 21 coding agent model and harness combinations to measure harness impact.", "origin": "backfill", "linkers": [ "harness-hub.org" ] }, { "id": "50021693", "url": "https://videodb.io/blog/a-faster-path-to-the-next-policy", "domain": "videodb.io", "title": "Action retrieval: VideoDB vs Nova and Twelve Labs", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50021693", "points": 2, "comments": 0, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "在100小时动作检索基准上对比VideoDB与竞品", "summary_en": "Vendor benchmark comparing action retrieval in 100 hours of video against AWS Nova and Twelve Labs.", "origin": "backfill", "linkers": [] }, { "id": "50022208", "url": "https://deepclause.substack.com/p/a-harness-in-a-single-file-agent", "domain": "deepclause.substack.com", "title": "A Harness in a single file: Agent skills as an evolving REPL module", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50022208", "points": 1, "comments": 0, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "讲解如何把agent技能做成单文件REPL模块", "summary_en": "Tutorial on packaging agent instructions, rules and memory into a single-file REPL harness.", "origin": "backfill", "linkers": [] }, { "id": "50023048", "url": "https://charlesxu.io/tp-cp-sp-ep-dp-pp/", "domain": "charlesxu.io", "title": "Trace a Token Through TP + CP + SP + EP + DP + PP in MoE Inference", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50023048", "points": 1, "comments": 0, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "讲解MoE推理中多种并行方式如何组合", "summary_en": "Walkthrough of how six parallelism schemes compose in MoE inference.", "origin": "backfill", "linkers": [] }, { "id": "50024121", "url": "https://s2lab.org/U-Space/", "domain": "s2lab.org", "title": "U-Space", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50024121", "points": 1, "comments": 0, "date": "2026-10-09", "submissions": 1, "source": "hn", "summary_zh": "研究语言模型不确定性产生的时机与原因", "summary_en": "Preprint project page studying when and why uncertainty arises in language models.", "origin": "backfill", "linkers": [] }, { "id": "50030180", "url": "https://aleph-alpha.com/en/blog/designing-kolibri-architecture-trade-offs-from-first-principles/", "domain": "aleph-alpha.com", "title": "Designing Kolibri: Architecture Trade-Offs from First Principles", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50030180", "points": 2, "comments": 0, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "分析Kolibri模型的架构取舍与参数分配", "summary_en": "Reports architecture trade-off analysis for the Kolibri model covering parameters, FLOPs and mixers.", "origin": "backfill", "linkers": [] }, { "id": "50030387", "url": "https://syfi.cs.washington.edu/blog/2026-10-08-qwen35-mi300a/", "domain": "syfi.cs.washington.edu", "title": "VibeSys Builds a Qwen3.5-397B Engine on MI300A, 2.3× Tuned SGLang", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50030387", "points": 2, "comments": 0, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "在MI300A上构建Qwen3.5-397B推理引擎，快于SGLang", "summary_en": "Reports a Qwen3.5-397B inference engine on MI300A running 2.3x faster than tuned SGLang.", "origin": "backfill", "linkers": [] }, { "id": "50032055", "url": "https://www.nandakishorm.com/writing/vega", "domain": "nandakishorm.com", "title": "Vega: a decision model that rolls a ball down a hill / Nandakishor M", "category": "research_report", "topic": "ai-ml", "confidence": "low", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50032055", "points": 1, "comments": 0, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "介绍决策模型Vega并说明其机制", "summary_en": "Describes Vega, a decision model, and how its pooled vectors are settled by a small physics engine.", "origin": "backfill", "linkers": [] }, { "id": "50032575", "url": "https://data4sci.com/blog/local-llm-inference-at-scale-with", "domain": "data4sci.com", "title": "Local LLM Inference at Scale with vLLM", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": true, "github_io": false, "cit": false, "hn_id": "50032575", "points": 1, "comments": 0, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "讲解用vLLM本地大规模运行LLM推理", "summary_en": "Tutorial on vLLM features such as continuous batching and prefix caching for local inference.", "origin": "backfill", "linkers": [] }, { "id": "50033374", "url": "https://machineplaying.org/", "domain": "machineplaying.org", "title": "Is this machine playing?", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50033374", "points": 1, "comments": 1, "date": "2026-10-10", "submissions": 2, "source": "hn", "summary_zh": "观察编码agent在无目标3D岛屿中的自发行为", "summary_en": "Reports observations of a coding agent left without a goal in a 3D island.", "origin": "backfill", "linkers": [] }, { "id": "50034182", "url": "https://pournasserian.com/writing/what-ai-hosts-require-of-an-mcp-server", "domain": "pournasserian.com", "title": "MCP server rules: Claude, ChatGPT, Codex, VS Code and more", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50034182", "points": 1, "comments": 0, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "比较八个AI宿主对MCP服务器的检查规则", "summary_en": "Compares checkable rules that eight AI hosts apply to MCP servers.", "origin": "backfill", "linkers": [] }, { "id": "50035357", "url": "https://breadbowl.ai/blog/breadbowl-embed/", "domain": "breadbowl.ai", "title": "One vector is too few. One per token is too many.", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": false, "hf": true, "github_io": false, "cit": false, "hn_id": "50035357", "points": 2, "comments": 1, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "提出用KV cache作嵌入的late-interaction架构", "summary_en": "Proposes BreadBowl-Embed, a late-interaction embedding architecture built on KV cache.", "origin": "backfill", "linkers": [] }, { "id": "50037203", "url": "https://epoch.ai/publications/the-plunging-price-of-thought", "domain": "epoch.ai", "title": "The plunging price of thought", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50037203", "points": 1, "comments": 0, "date": "2026-10-10", "submissions": 1, "source": "hn", "summary_zh": "测量AI性能成本下降速度并与其他技术比较", "summary_en": "Measures how fast the cost of a given AI performance level is falling versus other technologies.", "origin": "backfill", "linkers": [ "non.io", "stevekrouse.com", "substack.com", "transitivebullsh.it", "withnative.ai" ] }, { "id": "50040309", "url": "https://zrt.wtf/publication/2026-10-10-taste-route", "domain": "zrt.wtf", "title": "TasteRoute: Personalized Routing for Video Generation", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50040309", "points": 1, "comments": 0, "date": "2026-10-11", "submissions": 1, "source": "hn", "summary_zh": "提出个性化视频生成模型路由方法TasteRoute", "summary_en": "Presents TasteRoute, personalized routing among video generation models.", "origin": "backfill", "linkers": [] }, { "id": "50041006", "url": "https://notesbylex.com/building-an-agent-harness-from-scratch", "domain": "notesbylex.com", "title": "Building an agent harness from scratch to pass Ian Goodfellow's intelligence test", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50041006", "points": 2, "comments": 0, "date": "2026-10-11", "submissions": 2, "source": "hn", "summary_zh": "从零构建agent harness以通过Goodfellow测试", "summary_en": "Blog tutorial building an agent harness from scratch to pass a classic intelligence test.", "origin": "backfill", "linkers": [] }, { "id": "50042932", "url": "https://bitsofwork.org", "domain": "bitsofwork.org", "title": "Measuring Cryptanalytic Effort", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50042932", "points": 2, "comments": 0, "date": "2026-10-11", "submissions": 3, "source": "hn", "summary_zh": "用文献计量估算各公钥假设的密码分析投入", "summary_en": "Bibliometric estimate of public cryptanalysis effort for each public-key assumption.", "origin": "backfill", "linkers": [] }, { "id": "50048361", "url": "https://dreamstation.systems/personal/crypto30years.html", "domain": "dreamstation.systems", "title": "Has mainstream cryptography provided 30 years of protection? (Mostly not, but we’re getting way better)", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50048361", "points": 2, "comments": 0, "date": "2026-10-11", "submissions": 1, "source": "hn", "summary_zh": "回顾主流密码算法30年内是否被实际攻破", "summary_en": "Reviews whether widely used ciphers and hashes survived 30 years without practical attacks.", "origin": "backfill", "linkers": [] }, { "id": "50054201", "url": "https://www.drjoshcsimmons.com/writing/replaying-an-agent-should-not-call-the-model", "domain": "drjoshcsimmons.com", "title": "Replaying an agent should not call the model", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50054201", "points": 1, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "用可运行示例说明agent回放而不调用模型", "summary_en": "Runnable example explaining agent replay and request matching without calling the model.", "origin": "backfill", "linkers": [] }, { "id": "50056257", "url": "https://ceobench.com/", "domain": "ceobench.com", "title": "CEO-Bench", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50056257", "points": 2, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "提出CEO-Bench评测agent长期经营模拟公司", "summary_en": "Introduces CEO-Bench, evaluating whether AI agents can run a simulated startup for 500 days.", "origin": "backfill", "linkers": [] }, { "id": "50056329", "url": "https://www.construction-physics.com/p/understanding-the-ai-that-drives", "domain": "construction-physics.com", "title": "Understanding the AI That Drives Robots", "category": "explainer", "topic": "robotics-graphics", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50056329", "points": 2, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "讲解驱动机器人的视觉语言动作模型", "summary_en": "Explainer on vision-language-action models that drive robots.", "origin": "backfill", "linkers": [] }, { "id": "50057025", "url": "https://thecout.com/blog/modelsafety/", "domain": "thecout.com", "title": "What you actually download when you download a model", "category": "explainer", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50057025", "points": 1, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "讲解下载模型时的代码执行与权重后门风险", "summary_en": "Explains what is in a downloaded model, from executable files to weight backdoors.", "origin": "backfill", "linkers": [] }, { "id": "50059391", "url": "https://research.meta.ai/blog/solving-open-research-problems-together", "domain": "research.meta.ai", "title": "Solving Open Research Problems Together", "category": "research_report", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50059391", "points": 2, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "介绍数学家与Muse Spark合作完成六篇研究论文", "summary_en": "Meta reports mathematicians and its model collaborating on six research papers.", "origin": "backfill", "linkers": [] }, { "id": "50063679", "url": "https://www.asimov.press/p/dna", "domain": "asimov.press", "title": "Making DNA: The Quest for Synthetic Life", "category": "explainer", "topic": "bio-health", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50063679", "points": 2, "comments": 0, "date": "2026-10-12", "submissions": 1, "source": "hn", "summary_zh": "介绍合成基因组与合成生命的研究进展", "summary_en": "Feature explaining synthetic genomes, synthetic life and their biosecurity risks.", "origin": "backfill", "linkers": [] }, { "id": "50069793", "url": "https://www.naturewill.ai/blog/natura-1", "domain": "naturewill.ai", "title": "Natura-1: The Will to Act", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50069793", "points": 2, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "报告50B模型Natura-1用人类视频驱动20自由度灵巧手", "summary_en": "Technical report on Natura-1, a 50B model trained on human video driving a 20-DoF hand.", "origin": "backfill", "linkers": [] }, { "id": "50069926", "url": "https://www.nanoswe.com/blog.html", "domain": "nanoswe.com", "title": "blog - nanoswe", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50069926", "points": 1, "comments": 1, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "以1000美元预算从零训练agentic LLM并测SWE-bench", "summary_en": "Reports training an agentic coding LLM from scratch on a $1k budget, evaluated on SWE-bench.", "origin": "backfill", "linkers": [] }, { "id": "50071497", "url": "https://kuber.studio/planethunt/", "domain": "kuber.studio", "title": "KIC 5689121—A Kepler Planet Candidate", "category": "research_report", "topic": "physics-earth", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50071497", "points": 1, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "报告在Kepler数据中发现13.77天行星候选体", "summary_en": "Reports a 13.77-day planet candidate found in Kepler transit data with AI.", "origin": "backfill", "linkers": [] }, { "id": "50075655", "url": "https://www.amazon.science/blog/can-we-change-an-ai-models-architecture-without-retraining-it-from-scratch", "domain": "amazon.science", "title": "Can we change an AI model’s architecture without retraining it from scratch?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50075655", "points": 2, "comments": 1, "date": "2026-10-13", "submissions": 2, "source": "hn", "summary_zh": "提出Priming方法复用权重以改变模型架构", "summary_en": "Presents Priming, reusing pretrained Transformer weights to adopt more efficient architectures without full retraining.", "origin": "backfill", "linkers": [] }, { "id": "50079091", "url": "https://equistamp.substack.com/p/model-backdoors-are-easy-to-create", "domain": "equistamp.substack.com", "title": "Model backdoors are easy to create and hard to reliably remove", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50079091", "points": 2, "comments": 0, "date": "2026-10-13", "submissions": 1, "source": "hn", "summary_zh": "报告模型后门易植入而难以可靠移除", "summary_en": "Reports results that model backdoors are easy to insert and hard to reliably remove.", "origin": "backfill", "linkers": [ "norabble.com" ] }, { "id": "50088279", "url": "https://aistack.imec-int.com/blog/quantization", "domain": "aistack.imec-int.com", "title": "Optimization: does quantization hurt coding agents?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50088279", "points": 2, "comments": 1, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "评测量化对编码agent的影响", "summary_en": "Measures how FP8, INT4 AWQ and NVFP4 quantization affect coding agent performance.", "origin": "backfill", "linkers": [] }, { "id": "50090174", "url": "https://henniger.dev/post/djinn/", "domain": "henniger.dev", "title": "rust-djinn: You Should Specify Your Code Unambiguously (and Prove That You Did)", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50090174", "points": 2, "comments": 1, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "用300行规范实现并证明gzip解压器", "summary_en": "Presents a 300-line specified, proven gzip decompressor in Rust.", "origin": "backfill", "linkers": [] }, { "id": "50091404", "url": "https://llmknowledge.org/ch5", "domain": "llmknowledge.org", "title": "RAG Compliance: Where Confidential Data Actually Goes", "category": "explainer", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50091404", "points": 1, "comments": 0, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "讲解RAG系统中机密数据的流向与合规", "summary_en": "Explains where confidential data goes in RAG systems and compliance implications.", "origin": "backfill", "linkers": [] }, { "id": "50093745", "url": "https://scalingintelligence.stanford.edu/blogs/ncm/", "domain": "scalingintelligence.stanford.edu", "title": "Next-Concept Modeling: Fast, Parallel Coarse-to-Fine Text Generation | Scaling Intelligence Lab at Stanford University", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50093745", "points": 2, "comments": 0, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "提出Next-Concept Modeling并行粗到细文本生成", "summary_en": "Presents Next-Concept Modeling for parallel coarse-to-fine text generation.", "origin": "backfill", "linkers": [] }, { "id": "50094355", "url": "https://epoch.ai/publications/ai-systems-are-improving-rapidly-at-slay-the-spire-and-its-sequel", "domain": "epoch.ai", "title": "How well can AI play Slay the Spire?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50094355", "points": 2, "comments": 2, "date": "2026-10-14", "submissions": 2, "source": "hn", "summary_zh": "追踪前沿模型在Slay the Spire游戏中的进步", "summary_en": "Tracks frontier model win rates on Slay the Spire and its sequel.", "origin": "backfill", "linkers": [] }, { "id": "50096325", "url": "https://www.senseon.io/blog/netsupport-rat-analysis", "domain": "senseon.io", "title": "Into the Rat's Nest: A SenseOn Analysis of the...", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50096325", "points": 1, "comments": 0, "date": "2026-10-14", "submissions": 1, "source": "hn", "summary_zh": "技术分析NetSupport RAT的投递与持久化", "summary_en": "Technical analysis of NetSupport RAT campaigns covering delivery, persistence and command-and-control.", "origin": "backfill", "linkers": [] }, { "id": "50101303", "url": "https://three.dev/blog/knowledge-just-moved-out-of-the-gpu", "domain": "three.dev", "title": "Engram and N-gram memory in DeepSeek V4.1-Flash and Qwen3.8-Flash-Next: what changes for inference cost", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50101303", "points": 2, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "讲解Engram查找表架构及其脱离GPU的做法", "summary_en": "Overview of the Engram lookup-table architecture and why it can live off the GPU.", "origin": "backfill", "linkers": [] }, { "id": "50101568", "url": "https://knl.co.in/posts/gemma-3-1b-ternary/", "domain": "knl.co.in", "title": "Gemma 3 1B at 114 tokens per second on a laptop CPU", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50101568", "points": 2, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "在笔记本CPU上将Gemma 3 1B提速到114 tokens/s", "summary_en": "Reports ternary conversion and custom kernels getting Gemma 3 1B to 114 tokens per second on a laptop CPU.", "origin": "backfill", "linkers": [] }, { "id": "50103879", "url": "https://kernelspace.substack.com/p/building-a-gpu-compiler-from-scratch", "domain": "kernelspace.substack.com", "title": "Building a GPU Compiler from Scratch", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50103879", "points": 2, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "讲解从零构建GPU编译器的多级IR降低", "summary_en": "Tutorial on lowering PyTorch graphs to CUDA through multi-level IRs.", "origin": "backfill", "linkers": [] }, { "id": "50103886", "url": "https://confidential.ai/blog/kv-offload-under-cc", "domain": "confidential.ai", "title": "Confidential AI ･ Encrypted KV Cache Offloading for Confidential Computing", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50103886", "points": 1, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "提出面向机密计算的加密KV cache卸载", "summary_en": "Presents encrypted KV cache offloading for confidential computing as a production implementation.", "origin": "backfill", "linkers": [] }, { "id": "50104723", "url": "https://www.explainthatmove.com/", "domain": "explainthatmove.com", "title": "ExplainThatMove", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50104723", "points": 2, "comments": 1, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "比较并用自然语言解释国际象棋着法优劣", "summary_en": "Compares chess moves and explains in natural language why the stronger one is better.", "origin": "backfill", "linkers": [] }, { "id": "50105571", "url": "https://www.dbreunig.com/2026/10/14/context-inertia.html", "domain": "dbreunig.com", "title": "Context Inertia: When Agents Won’t Let Go", "category": "research_report", "topic": "ai-ml", "confidence": "low", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50105571", "points": 1, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "提出并讨论agent的Context Inertia现象", "summary_en": "Describes context inertia, where agents keep pursuing goals despite contrary evidence.", "origin": "backfill", "linkers": [] }, { "id": "50107596", "url": "https://gimletlabs.ai/blog/inside-agentic-ai-workloads", "domain": "gimletlabs.ai", "title": "Inside Agentic AI Workloads", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50107596", "points": 2, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "刻画agentic AI工作负载的推理特征", "summary_en": "Characterizes agentic AI workloads that repeatedly re-enter model inference over growing state.", "origin": "backfill", "linkers": [] }, { "id": "50110143", "url": "https://www.vals.ai/blogs/routers-new-ai-primitive", "domain": "vals.ai", "title": "Routers Are a New AI Primitive", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50110143", "points": 2, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "评测模型路由器的质量、成本与延迟", "summary_en": "Argues for and runs independent evaluation of model routers on quality, cost and latency.", "origin": "backfill", "linkers": [] }, { "id": "50110814", "url": "https://rlafuente.com/posts/2026-10-11-agent-evals-4-verifiers-first-rubrics-second.html", "domain": "rlafuente.com", "title": "Rodney L.", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50110814", "points": 1, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "讲解LLM agent评测中先用程序验证再用rubric打分", "summary_en": "Explains why programmatic verifiers should come before rubrics when grading LLM agents.", "origin": "backfill", "linkers": [] }, { "id": "50111102", "url": "https://www.lesswrong.com/posts/c9woewdY9ddsPqf3J/the-continual-constitutional-convention-how-language-models", "domain": "lesswrong.com", "title": "The Continual Constitutional Convention: How language models revise the documents that govern them", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50111102", "points": 1, "comments": 0, "date": "2026-10-15", "submissions": 1, "source": "hn", "summary_zh": "研究语言模型如何修订约束自身的文件(宪法)", "summary_en": "Experiments studying how language models revise the constitutions that govern them.", "origin": "backfill", "linkers": [] }, { "id": "50116630", "url": "https://www.goodfire.com/research/reward-hacking-activation-monitors", "domain": "goodfire.com", "title": "Models know when they’re reward hacking—and we can catch them at scale", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50116630", "points": 2, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "发现模型内部与reward hacking对应的信号并训练probe", "summary_en": "Finds an internal signal accompanying reward hacking and builds probes to detect it.", "origin": "backfill", "linkers": [] }, { "id": "50116891", "url": "https://blog.neels.tech/blog/this-is-not-a-game-of-snake/", "domain": "blog.neels.tech", "title": "This is not a game of snake", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50116891", "points": 1, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "实验训练transformer学习玩贪吃蛇", "summary_en": "Experiment teaching a transformer model to play Snake and examining what it learns.", "origin": "backfill", "linkers": [] }, { "id": "50117824", "url": "https://modal.com/blog/toward-rl-on-mixed-precision-mixed-accelerator-samplers", "domain": "modal.com", "title": "Toward RL on mixed-precision, mixed-accelerator samplers", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50117824", "points": 2, "comments": 0, "date": "2026-10-16", "submissions": 4, "source": "hn", "summary_zh": "用混合精度混合加速器采样器做异步RL训练", "summary_en": "Reports asynchronous RL training for agentic coding using mixed-precision, mixed-accelerator samplers.", "origin": "backfill", "linkers": [] }, { "id": "50120094", "url": "https://bonsai.io/blog/pink-elephant-in-your-prompt/", "domain": "bonsai.io", "title": "Bonsai Blog | Fully Managed Elasticsearch & OpenSearch • The Pink Elephant in Your Prompt", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50120094", "points": 1, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "用720次试验检验让agent忽略已有代码的提示效果", "summary_en": "Reports 720 trials testing whether telling an agent to ignore existing code works.", "origin": "backfill", "linkers": [] }, { "id": "50120918", "url": "https://richardli.xyz/post/gradient-is-covariance/", "domain": "richardli.xyz", "title": "The Gradient Is a Covariance", "category": "explainer", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50120918", "points": 2, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "推导后训练更新中梯度与协方差的关系", "summary_en": "Derives the relation between policy gradient and covariance in post-training updates.", "origin": "backfill", "linkers": [] }, { "id": "50121190", "url": "https://nostalgia.dev/posts/running-doom-inside-a-transformer/", "domain": "nostalgia.dev", "title": "Running DOOM Inside a Transformer", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50121190", "points": 2, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "构建能运行DOOM的transformer", "summary_en": "Builds a transformer that executes the DOOM game and describes the method.", "origin": "backfill", "linkers": [] }, { "id": "50121458", "url": "https://biologyinsights.com/akkermansia-and-cancer-a-look-at-the-research/", "domain": "biologyinsights.com", "title": "Akkermansia and Cancer: A Look at the Research", "category": "explainer", "topic": "bio-health", "confidence": "medium", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50121458", "points": 1, "comments": 1, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "综述Akkermansia与癌症的研究", "summary_en": "Overview of research linking the gut microbe Akkermansia to cancer.", "origin": "backfill", "linkers": [] }, { "id": "50121608", "url": "https://llemre.com/benchmaxxing/", "domain": "llemre.com", "title": "A Statistical Analysis of LLM Benchmaxxing", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50121608", "points": 1, "comments": 1, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "统计分析各实验室模型是否在基准上过拟合", "summary_en": "Statistical estimate of benchmark overfitting by model lab.", "origin": "backfill", "linkers": [] }, { "id": "50121855", "url": "https://kyutai.org/reki/", "domain": "kyutai.org", "title": "REKI: Registering Knowledge in VLAs", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50121855", "points": 1, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "介绍带近期记忆的2.4B视觉语言动作模型REKI", "summary_en": "Presents REKI, a 2.4B vision-language-action model with memory, trained on public data.", "origin": "backfill", "linkers": [] }, { "id": "50124514", "url": "https://semiengineering.com/benchmarking-open-source-llms-on-verilog-rtl-generation-across-50-tasks-nmims-iit-roorkee-bits-pilani/", "domain": "semiengineering.com", "title": "Benchmarking Open-Source LLMs on Verilog RTL Generation Across 50 Tasks (NMIMS, IIT Roorkee, BITS Pilani)", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50124514", "points": 1, "comments": 0, "date": "2026-10-16", "submissions": 1, "source": "hn", "summary_zh": "报道在50个任务上评测开源LLM生成Verilog的论文", "summary_en": "News summary of a paper benchmarking open-source LLMs on Verilog RTL generation.", "origin": "backfill", "linkers": [] }, { "id": "50127044", "url": "https://alexwang.ai/posts/derivatives-and-gradients/", "domain": "alexwang.ai", "title": "Matrix gradients, without the pain", "category": "explainer", "topic": "math-theory", "confidence": "high", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50127044", "points": 2, "comments": 0, "date": "2026-10-17", "submissions": 1, "source": "hn", "summary_zh": "介绍简化矩阵梯度推导的技巧", "summary_en": "Teaches a simple trick for deriving matrix gradients without index notation.", "origin": "backfill", "linkers": [] }, { "id": "50131429", "url": "https://research.google/pubs/open-and-emergent-problems-in-agentic-privacy-and-security-a-contextual-angle/", "domain": "research.google", "title": "Open and Emergent Problems in Agentic Privacy and Security: A Contextual Angle", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50131429", "points": 1, "comments": 0, "date": "2026-10-17", "submissions": 1, "source": "hn", "summary_zh": "梳理agent隐私与安全的开放问题", "summary_en": "Technical report outlining open problems in agentic privacy and security.", "origin": "backfill", "linkers": [] }, { "id": "50133211", "url": "https://chesscheatdetector.com/blog/building-a-chess-cheat-detector", "domain": "chesscheatdetector.com", "title": "Building a Chess Cheat Detector", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50133211", "points": 2, "comments": 0, "date": "2026-10-17", "submissions": 1, "source": "hn", "summary_zh": "用Maia模拟6613局棋构建作弊检测器", "summary_en": "Builds a chess cheat detector from 6,613 simulated games with Maia.", "origin": "backfill", "linkers": [] }, { "id": "50133714", "url": "https://tileward.com/research/a-smaller-speech-recogniser/", "domain": "tileward.com", "title": "We shrank Whisper large-v3 by 68%, below the size of turbo", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50133714", "points": 2, "comments": 0, "date": "2026-10-17", "submissions": 1, "source": "hn", "summary_zh": "将Whisper large-v3压缩68%并报告测试结果", "summary_en": "Reports compressing Whisper large-v3 by 68% with accuracy results on ten tests.", "origin": "backfill", "linkers": [] }, { "id": "50137493", "url": "https://www.goodfire.com/research/production-cyber-monitors-on-kimi-k3", "domain": "goodfire.com", "title": "Training and Deploying Production Cyber Monitors on Kimi K3", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50137493", "points": 2, "comments": 0, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "为Kimi K3训练网络安全监控器并与LLM judge比较", "summary_en": "Trains cybersecurity monitors for Kimi K3 that beat an LLM judge at lower cost.", "origin": "backfill", "linkers": [] }, { "id": "50138777", "url": "https://amongbytes.com/posts/rfc-10024-negotiating-pq-tls-handshakes/", "domain": "amongbytes.com", "title": "RFC 10024: Negotiating Post-Quantum TLS handshakes", "category": "explainer", "topic": "security-privacy", "confidence": "medium", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50138777", "points": 2, "comments": 0, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "讲解后量子TLS握手协商与RFC 10024", "summary_en": "Explains post-quantum TLS handshake negotiation from the RFC 10024 co-author's view.", "origin": "backfill", "linkers": [] }, { "id": "50139703", "url": "https://workflows-of-ai.com/blog/can-ai-set-the-focal-point-of-a-photo", "domain": "workflows-of-ai.com", "title": "Can AI set the focal point of a photo?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50139703", "points": 1, "comments": 0, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "评测19种用LLM和模型裁剪图片的方法", "summary_en": "Evaluates 19 methods for cropping images using LLMs and classification models.", "origin": "backfill", "linkers": [] }, { "id": "50140159", "url": "https://www.benjoffe.com/fast-time-of-day", "domain": "benjoffe.com", "title": "A faster way to convert a timestamp ➜ Hour, Min, Sec", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50140159", "points": 2, "comments": 0, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "提出更快的时间戳转时分秒算法", "summary_en": "Presents a faster timestamp to hour-minute-second algorithm with benchmark results.", "origin": "backfill", "linkers": [] }, { "id": "50142691", "url": "https://nathanho.me/posts/evograin/", "domain": "nathanho.me", "title": "EvoGrain: Algorithmic Sound Design with L-Systems and Granular Synthesis", "category": "explainer", "topic": "other", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50142691", "points": 2, "comments": 0, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "介绍用L系统和颗粒合成做声音设计", "summary_en": "Describes algorithmic sound design using L-systems and granular synthesis.", "origin": "backfill", "linkers": [] }, { "id": "50142827", "url": "https://www.philipzucker.com/asmc_cpp/", "domain": "philipzucker.com", "title": "Ghidra + CBMC = Assembly Model Checker", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50142827", "points": 2, "comments": 0, "date": "2026-10-18", "submissions": 2, "source": "hn", "summary_zh": "将Ghidra汇编语义接入CBMC做模型检查", "summary_en": "Connects Ghidra lifted assembly semantics to CBMC to build an assembly model checker.", "origin": "backfill", "linkers": [] }, { "id": "50143274", "url": "https://adlrocha.substack.com/p/adlrocha-decision-models-as-a-judge", "domain": "adlrocha.substack.com", "title": "@adlrocha", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50143274", "points": 1, "comments": 0, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "介绍用decision model在CI中做agent评测", "summary_en": "Explains using decision models as judges to run agent evals in CI.", "origin": "backfill", "linkers": [] }, { "id": "50144489", "url": "https://warlock.ai/blog/warcraftsim-1/", "domain": "warlock.ai", "title": "Beating Warcraft III’s insane AI with self-play reinforcement learning on one desktop", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50144489", "points": 2, "comments": 0, "date": "2026-10-18", "submissions": 1, "source": "hn", "summary_zh": "用自对弈RL在单机上击败魔兽争霸3 AI", "summary_en": "Trains a self-play RL agent reaching 89% win rate against Warcraft III's built-in AI.", "origin": "backfill", "linkers": [] }, { "id": "50149314", "url": "https://localai-brief.vercel.app/", "domain": "localai-brief.vercel.app", "title": "How much intelligence can you run locally? (October 2026)", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50149314", "points": 1, "comments": 2, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "对比本地可运行模型与前沿模型的研究简报", "summary_en": "Research brief comparing locally runnable models with frontier models on benchmark indices.", "origin": "backfill", "linkers": [] }, { "id": "50149698", "url": "https://aiskills402.com/blog/agent-skills-security-risks", "domain": "aiskills402.com", "title": "Why you should not install unvetted AI agent skills", "category": "explainer", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50149698", "points": 1, "comments": 1, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "说明安装未审查agent skill的风险", "summary_en": "Explains security risks of installing unvetted AI agent skills.", "origin": "backfill", "linkers": [] }, { "id": "50151572", "url": "https://www.csis.org/analysis/art-swarming", "domain": "csis.org", "title": "The Art of Swarming", "category": "research_report", "topic": "other", "confidence": "medium", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50151572", "points": 2, "comments": 0, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "报告论证无人机集群改变战场集结方式", "summary_en": "Policy report arguing drone swarms change mass on the battlefield.", "origin": "backfill", "linkers": [] }, { "id": "50151774", "url": "https://ztatlock.net/pubs/2026-oopsla-eggcc/", "domain": "ztatlock.net", "title": "Efficient Extraction for Effectful E-graphs", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50151774", "points": 1, "comments": 0, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "提出Statewalk DP加速带副作用e-graph提取", "summary_en": "Presents Statewalk DP for fast effect-aware e-graph extraction without ILP.", "origin": "backfill", "linkers": [] }, { "id": "50152788", "url": "https://getjavo.com/research/sales-evaluation/", "domain": "getjavo.com", "title": "Evaluating Javo", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50152788", "points": 1, "comments": 0, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "在100个网站上评测AI销售agent", "summary_en": "Evaluates an AI sales agent on 100 websites and 14.8 million visitors.", "origin": "backfill", "linkers": [] }, { "id": "50154153", "url": "https://wiki.alcidesfonseca.com/blog/lessons-learned-building-fronteira/", "domain": "wiki.alcidesfonseca.com", "title": "Lessons learned building Fronteira by Alcides Fonseca", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50154153", "points": 2, "comments": 0, "date": "2026-10-19", "submissions": 2, "source": "hn", "summary_zh": "介绍用信息流策略防御的MCP代理Fronteira", "summary_en": "Describes Fronteira, an MCP proxy using information-flow policies to block attacks.", "origin": "backfill", "linkers": [] }, { "id": "50154484", "url": "https://fremaconsulting.ch/blog/backpropagation", "domain": "fremaconsulting.ch", "title": "Backpropagation Step by Step: The Forward and Backward Pass", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50154484", "points": 2, "comments": 1, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "手算演示反向传播的前向与反向过程", "summary_en": "Step-by-step hand-checked walkthrough of backpropagation in a small network.", "origin": "backfill", "linkers": [] }, { "id": "50154830", "url": "https://www.ultracite.ai/research/type-escape-hatches", "domain": "ultracite.ai", "title": "Why TypeScript lets you overrule its type checker", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50154830", "points": 1, "comments": 0, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "解释TypeScript中any与ts-ignore的来历", "summary_en": "Explains the origin and limits of escape hatches like any and ts-ignore in TypeScript.", "origin": "backfill", "linkers": [] }, { "id": "50155677", "url": "https://kapteko.com/running-10-million-codex-clis-on-one-server-lfi-processes/", "domain": "kapteko.com", "title": "Running 10 million Codex CLIs on one server-LFI processes", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50155677", "points": 1, "comments": 0, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "在单台服务器上运行千万级Codex CLI进程", "summary_en": "Describes running millions of Codex CLI processes on one server.", "origin": "backfill", "linkers": [] }, { "id": "50158847", "url": "https://www.lei.chat/posts/mlir-codegen-dialects-for-machine-learning-compilers/", "domain": "lei.chat", "title": "MLIR CodeGen Dialects for Machine Learning Compilers", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50158847", "points": 1, "comments": 1, "date": "2026-10-19", "submissions": 1, "source": "hn", "summary_zh": "介绍面向ML编译器的MLIR代码生成方言", "summary_en": "Explains MLIR codegen dialects for machine learning compilers.", "origin": "backfill", "linkers": [] }, { "id": "50162371", "url": "https://lumalabs.ai/news/packed-multimodal-attention", "domain": "lumalabs.ai", "title": "MARBLE: exact sequence parallelism for packed multimodal masks", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50162371", "points": 1, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "提出多模态packed序列的精确序列并行方法Marble", "summary_en": "Presents Marble, exact sequence parallelism for packed multimodal sequences, with results.", "origin": "backfill", "linkers": [] }, { "id": "50162758", "url": "https://hamel.dev/blog/posts/decision-model-evals/", "domain": "hamel.dev", "title": "How To Use Decision Models for Evals", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50162758", "points": 1, "comments": 0, "date": "2026-10-20", "submissions": 2, "source": "hn", "summary_zh": "讲解如何选择并调优evals的decision model", "summary_en": "Guide to choosing and tuning decision models for evals.", "origin": "backfill", "linkers": [] }, { "id": "50165645", "url": "https://www.stackone.com/blog/system-one-model-prompt-injection/", "domain": "stackone.com", "title": "Building a System One Model for Prompt Injection", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50165645", "points": 2, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "重建提示注入防护模型Defender并报告效果", "summary_en": "Describes rebuilding the Defender prompt injection guard with evaluation results.", "origin": "backfill", "linkers": [] }, { "id": "50165802", "url": "https://bart.degoe.de/fine-tuning-a-small-language-model-for-rag/", "domain": "bart.degoe.de", "title": "Fine-tuning a small language model for RAG in a browser on a Mac", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50165802", "points": 1, "comments": 1, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "在Mac上微调Qwen3-0.6B用于浏览器RAG", "summary_en": "Fine-tunes Qwen3-0.6B on a Mac for in-browser RAG and reports results.", "origin": "backfill", "linkers": [] }, { "id": "50167109", "url": "https://rocm.blogs.amd.com/software-tools-optimization/rdna-fp16-gemm/README.html", "domain": "rocm.blogs.amd.com", "title": "FP16 GEMM Optimization on AMD RDNA™ 3 and RDNA™ 4 Architectures", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50167109", "points": 2, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "逐步优化AMD RDNA上的FP16 GEMM内核", "summary_en": "Step-by-step optimization of an FP16 GEMM kernel on AMD RDNA 3 and 4.", "origin": "backfill", "linkers": [] }, { "id": "50167932", "url": "https://pentad.ai/PLRN/031/", "domain": "pentad.ai", "title": "Jev models make agents faster. Hooked control planes steal the difference.", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50167932", "points": 2, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "分析hook链对decision model加速收益的损耗", "summary_en": "Analyzes how hook chains absorb speedups from fast decision models.", "origin": "backfill", "linkers": [] }, { "id": "50168680", "url": "https://three.dev/blog/the-free-upgrade-that-raised-your-bill", "domain": "three.dev", "title": "The free upgrade that raised your bill: what a new model version really costs", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50168680", "points": 2, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "测量新模型版本如何改变实际API账单", "summary_en": "Measures how new model versions change real API bills at equal list price.", "origin": "backfill", "linkers": [] }, { "id": "50169048", "url": "https://alignment.openai.com/beyond-gmean-squared/", "domain": "alignment.openai.com", "title": "Beyond g-mean²: complementary metrics for monitorability", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50169048", "points": 1, "comments": 0, "date": "2026-10-20", "submissions": 2, "source": "hn", "summary_zh": "提出两个补充g-mean²的可监控性指标", "summary_en": "Introduces two complementary monitorability metrics beyond g-mean².", "origin": "backfill", "linkers": [] }, { "id": "50169298", "url": "https://getswiftapi.com", "domain": "getswiftapi.com", "title": "Alignment Is a System Property, Not Just a Model Property", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": true, "github_io": false, "cit": false, "hn_id": "50169298", "points": 2, "comments": 0, "date": "2026-10-13", "submissions": 3, "source": "hn", "summary_zh": "报告GPT-4-0613系统提示消融实验", "summary_en": "Reports a system-prompt ablation on GPT-4-0613 finding zero-byte responses.", "origin": "backfill", "linkers": [] }, { "id": "50170679", "url": "https://magiroux.com/posts/from-use-case-to-graphql-operation", "domain": "magiroux.com", "title": "From Use Case to GraphQL Operation: A Layered Approach for Agents", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50170679", "points": 1, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "介绍agent从用例到GraphQL操作的分层方法", "summary_en": "Describes a layered approach for agents to turn use cases into GraphQL operations.", "origin": "backfill", "linkers": [] }, { "id": "50171558", "url": "https://www.journalovi.org/2024-Cashman-PAC-learning-game/", "domain": "journalovi.org", "title": "PAC Learning; Or: Why We Should (and Shouldn't) Trust Machine Learning", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50171558", "points": 2, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "用交互游戏说明PAC学习", "summary_en": "Interactive game illustrating PAC learning and why ML models fail in deployment.", "origin": "backfill", "linkers": [] }, { "id": "50172326", "url": "https://nevolodia.com/thoughts/001/", "domain": "nevolodia.com", "title": "RELAGRAM", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50172326", "points": 1, "comments": 0, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "介绍用于车辆路径的潜变量递归RL方法Relagram", "summary_en": "Describes Relagram, latent recursion RL for vehicle routing.", "origin": "backfill", "linkers": [] }, { "id": "50173269", "url": "https://saulius.io/blog/what-are-looped-transformers-recurrent-depth", "domain": "saulius.io", "title": "What Are Looped Transformers? A Technical Deep Dive", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50173269", "points": 1, "comments": 1, "date": "2026-10-20", "submissions": 1, "source": "hn", "summary_zh": "技术讲解looped transformer", "summary_en": "Technical explanation of looped transformers.", "origin": "backfill", "linkers": [] }, { "id": "50175574", "url": "https://mulgadc.com/blog/blueprints-not-prompts-tla-plus", "domain": "mulgadc.com", "title": "Blueprints, Not Prompts: TLA+ for AI-Written Code", "category": "explainer", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50175574", "points": 2, "comments": 1, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "介绍用TLA+指导AI编写代码", "summary_en": "Describes using TLA+ specifications to guide AI-written code.", "origin": "backfill", "linkers": [] }, { "id": "50175763", "url": "https://www.spacebar.ai/benchmark", "domain": "spacebar.ai", "title": "AI Agents in a Shared Workspace: Benchmark", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50175763", "points": 1, "comments": 0, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "提出评测agent在共享工作区表现的SpacebarBench", "summary_en": "Introduces SpacebarBench for evaluating agents in a shared workspace.", "origin": "backfill", "linkers": [] }, { "id": "50176035", "url": "https://pantheon.inc/research/diverse-umi", "domain": "pantheon.inc", "title": "Pantheon | Building the Most Diverse UMI Dataset in Robotics", "category": "research_report", "topic": "robotics-graphics", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50176035", "points": 1, "comments": 0, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "介绍任务多样的UMI机器人数据集", "summary_en": "Describes a diverse UMI robotics dataset collection operation.", "origin": "backfill", "linkers": [] }, { "id": "50180067", "url": "https://rfd4-proteina.org/", "domain": "rfd4-proteina.org", "title": "RFD4-Proteina", "category": "research_report", "topic": "bio-health", "confidence": "high", "interactive": true, "repo": true, "arxiv": false, "hf": true, "github_io": false, "cit": false, "hn_id": "50180067", "points": 1, "comments": 0, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "提出生物分子设计模型RFD4-Proteina", "summary_en": "Presents RFD4-Proteina, a latent reasoning model for biomolecular design.", "origin": "backfill", "linkers": [] }, { "id": "50181579", "url": "https://www.ctgt.ai/research/concordance", "domain": "ctgt.ai", "title": "Deterministic Inference Did Not Make Our Pipeline Deterministic", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50181579", "points": 2, "comments": 0, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "实验显示确定性推理未使流水线确定", "summary_en": "Experiment showing deterministic LLM inference did not make a pipeline deterministic.", "origin": "backfill", "linkers": [] }, { "id": "50185061", "url": "https://blog.gopfert.eu/ai-macros-from-photos", "domain": "blog.gopfert.eu", "title": "The Overfit", "category": "research_report", "topic": "bio-health", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50185061", "points": 1, "comments": 1, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "称重32餐评估AI照片估算营养的准确性", "summary_en": "Weighs 32 meals to test how accurately AI estimates macros from photos.", "origin": "backfill", "linkers": [] }, { "id": "50187401", "url": "https://curtjaimungal.substack.com/p/how-witten-used-quantum-field-theory-105", "domain": "curtjaimungal.substack.com", "title": "How Witten Used Quantum Field Theory to Understand Knots", "category": "explainer", "topic": "physics-earth", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50187401", "points": 2, "comments": 0, "date": "2026-10-21", "submissions": 1, "source": "hn", "summary_zh": "讲解Witten如何用量子场论研究纽结", "summary_en": "Explains how Witten used quantum field theory to study knots.", "origin": "backfill", "linkers": [] }, { "id": "50191079", "url": "https://natolambert.com/slides/2026/harvard/lecture.html", "domain": "natolambert.com", "title": "Open Language Models", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50191079", "points": 1, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "开放语言模型的客座讲座幻灯片", "summary_en": "Lecture slides on open language models.", "origin": "backfill", "linkers": [] }, { "id": "50191260", "url": "https://topos.institute/blog/2026-10-14-e-graphs-3/", "domain": "topos.institute", "title": "How to prove equations using diagrams, Part 3", "category": "explainer", "topic": "math-theory", "confidence": "high", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50191260", "points": 1, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "讲解用图示证明等式的范畴语义", "summary_en": "Explains diagrammatic equational reasoning using initial functor semantics.", "origin": "backfill", "linkers": [] }, { "id": "50193915", "url": "https://secondtimeisabug.com/blog/what-projects-tell-ai-coding-agents/", "domain": "secondtimeisabug.com", "title": "What 1,866 GitHub Projects Tell Their AI Coding Agents", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50193915", "points": 1, "comments": 0, "date": "2026-10-15", "submissions": 2, "source": "hn", "summary_zh": "分析1866个GitHub项目的AGENTS.md", "summary_en": "Analyzes AGENTS.md files from 1,866 popular GitHub projects.", "origin": "backfill", "linkers": [] }, { "id": "50195326", "url": "https://www.hugo.im/posts/code-review-in-one-year/", "domain": "hugo.im", "title": "Machines Read, Humans Sign: Code Review in One Year", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50195326", "points": 1, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "总结自建代码审查agent一年的数据", "summary_en": "Reports one year of data from an in-house code review agent.", "origin": "backfill", "linkers": [] }, { "id": "50195366", "url": "https://subway.bio1up.com/line/signal", "domain": "subway.bio1up.com", "title": "How GLP-1 drugs like semaglutide work", "category": "explainer", "topic": "bio-health", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50195366", "points": 1, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "用地铁隐喻讲解GLP-1作用机制", "summary_en": "Interactive subway-themed explainer of how GLP-1 works.", "origin": "backfill", "linkers": [] }, { "id": "50195427", "url": "https://counterparts.ai/ecosystem/", "domain": "counterparts.ai", "title": "The Memory Field Guide", "category": "explainer", "topic": "bio-health", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50195427", "points": 2, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "人类与人工记忆的交互式指南", "summary_en": "Interactive guide to human and artificial memory.", "origin": "backfill", "linkers": [] }, { "id": "50195661", "url": "https://www.bis.org/publications/working-paper-1350-elasticity-money-production-networks-working-capital-credit-lines-and-financial-conditions", "domain": "bis.org", "title": "Elasticity of money in production networks, working capital, credit lines and financial conditions", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50195661", "points": 1, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "研究生产网络中货币弹性", "summary_en": "Research paper on money elasticity, credit lines and production networks.", "origin": "backfill", "linkers": [] }, { "id": "50196194", "url": "https://www.microsoft.com/en-us/research/blog/solving-a-31-year-old-grand-challenge-in-mathematics-the-kls-conjecture/", "domain": "microsoft.com", "title": "Solving a 31-year-old grand challenge in mathematics–the KLS conjecture", "category": "research_report", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50196194", "points": 2, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "宣布借助AI解决KLS猜想", "summary_en": "Announces resolution of the KLS conjecture with AI assistance.", "origin": "backfill", "linkers": [] }, { "id": "50196328", "url": "https://rhg.com/research/chinas-next-generation-industrial-policy/", "domain": "rhg.com", "title": "China’s Next-Generation Industrial Policy", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50196328", "points": 1, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "分析中国新一代产业政策的报告", "summary_en": "Report analyzing China's next-generation industrial policy.", "origin": "backfill", "linkers": [] }, { "id": "50198515", "url": "https://harness-hacks.vercel.app", "domain": "harness-hacks.vercel.app", "title": "A history of Harness Hacks", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50198515", "points": 2, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "讲述prompt harness技巧的历史", "summary_en": "Slide talk on the history of harness hacks for prompting.", "origin": "backfill", "linkers": [] }, { "id": "50201241", "url": "https://labs.baseten.co/articles/mammoth-1-and-elephant-1-frontier-decision-models", "domain": "labs.baseten.co", "title": "Mammoth-1 and Elephant-1: frontier decision models", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50201241", "points": 1, "comments": 0, "date": "2026-10-22", "submissions": 1, "source": "hn", "summary_zh": "介绍decision model Mammoth-1与Elephant-1", "summary_en": "Technical report on the Mammoth-1 and Elephant-1 decision models.", "origin": "backfill", "linkers": [] }, { "id": "50203118", "url": "https://www.lesswrong.com/posts/ozbbzsh49wvgfbCDp/twitch-plays-ai-interpretability-of-bonsai-8b", "domain": "lesswrong.com", "title": "Twitch Plays: AI Interpretability of Bonsai-8B", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50203118", "points": 1, "comments": 1, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "直播式探索Bonsai-8B可解释性", "summary_en": "Live interpretability exploration of Bonsai-8B with a public interpretability map.", "origin": "backfill", "linkers": [] }, { "id": "50204024", "url": "https://malwarebench.com", "domain": "malwarebench.com", "title": "MalwareBench: Reverse-engineering custom-built, realistic, evasive malware into actionable intelligence", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50204024", "points": 1, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "MalwareBench评测前沿模型逆向分析恶意软件的能力", "summary_en": "Evaluates frontier models on reverse engineering evasive malware across platforms with a new benchmark.", "origin": "backfill", "linkers": [] }, { "id": "50204050", "url": "https://frontier-cs.org/?v=2", "domain": "frontier-cs.org", "title": "FrontierCS 2", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50204050", "points": 1, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "FrontierCS 2基准用48个开放式CS研发任务评测AI智能体", "summary_en": "A benchmark of 48 open-ended computer science R&D tasks for evaluating AI agents.", "origin": "backfill", "linkers": [] }, { "id": "50205746", "url": "https://finitering.space/38-s13/lab.html", "domain": "finitering.space", "title": "Finite Ring Cosmology Lab: Schrödinger—Dirac dynamics of the total state over the (13, 233) holographic substrate", "category": "research_report", "topic": "physics-earth", "confidence": "low", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50205746", "points": 1, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "用浏览器模拟Schrödinger-Dirac动力学的全息宇宙模型", "summary_en": "Presents a browser simulation of Schrödinger-Dirac dynamics on a holographic substrate with a claimed universe model.", "origin": "backfill", "linkers": [] }, { "id": "50206276", "url": "https://knowledgepixels.com/nanopub-ecosystem-paper/", "domain": "knowledgepixels.com", "title": "A Robust Decentralized Infrastructure for Trust-Aware Open Knowledge Sharing", "category": "research_report", "topic": "scholarly-communication", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50206276", "points": 2, "comments": 5, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "提出去中心化、可信感知的开放知识共享基础设施", "summary_en": "Describes a decentralized, trust-aware infrastructure for open knowledge sharing.", "origin": "backfill", "linkers": [] }, { "id": "50207332", "url": "https://zerowithdot.com/blog/text-to-diagram-for-ai-systems/", "domain": "zerowithdot.com", "title": "Text-to-Diagram for AI Systems: Designing a DSL an LLM Gets Right", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50207332", "points": 1, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "介绍为LLM设计的图表DSL Gnomon及其编译器", "summary_en": "Describes Gnomon, a DSL and Python compiler for generating diagrams that an LLM can write reliably.", "origin": "backfill", "linkers": [] }, { "id": "50208384", "url": "https://gking.harvard.edu/peerless/", "domain": "gking.harvard.edu", "title": "AI Article Reviews Dominate Human Peer Reviews at Their Best: Evidence and Policy Choices", "category": "research_report", "topic": "scholarly-communication", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50208384", "points": 2, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "实证显示AI论文评审在多项指标上优于人类评审", "summary_en": "Presents evidence that AI-generated article reviews outperform human peer reviews on many measures, with policy implications.", "origin": "backfill", "linkers": [] }, { "id": "50209358", "url": "https://hypha.pub/the-language-that-says-no", "domain": "hypha.pub", "title": "The language that says no", "category": "explainer", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50209358", "points": 1, "comments": 0, "date": "2026-10-22", "submissions": 2, "source": "hn", "summary_zh": "从Milner的LCF讲到智能体用OCaml写代码", "summary_en": "Traces the history of Milner's LCF and argues for type-safe languages like OCaml for coding agents.", "origin": "backfill", "linkers": [] }, { "id": "50211931", "url": "https://yamanahlawat.in/articles/llm-judge-same-code-different-scores", "domain": "yamanahlawat.in", "title": "An LLM judge passed renders with defects I planted", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50211931", "points": 1, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "实验显示视觉LLM评审会放过人为植入缺陷的渲染", "summary_en": "Experiments show a vision-model judge passes renders with planted defects and is inconsistent across runs.", "origin": "backfill", "linkers": [] }, { "id": "50212738", "url": "https://www.gilesthomas.com/2026/10/rmsnorm", "domain": "gilesthomas.com", "title": "RMSNorm :: Giles' blog", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50212738", "points": 2, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "讲解RMSNorm及其相对LayerNorm的特点", "summary_en": "A tutorial explaining RMSNorm and how it compares with LayerNorm.", "origin": "backfill", "linkers": [] }, { "id": "50215711", "url": "https://jakubcwirlej.com/ai-dev/", "domain": "jakubcwirlej.com", "title": "Jakub Ćwirlej", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50215711", "points": 1, "comments": 0, "date": "2026-10-23", "submissions": 1, "source": "hn", "summary_zh": "单GPU训练160M代码模型,HumanEval达22.56%", "summary_en": "Documents training a 160M code model on one GPU that reaches 22.56% on HumanEval.", "origin": "backfill", "linkers": [] }, { "id": "50219307", "url": "https://huggingface.co/blog/Luo-Yihang/mundo-world-model", "domain": "huggingface.co", "title": "Mundo: From seeing a place to simulating what happens there", "category": "research_report", "topic": "robotics-graphics", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50219307", "points": 1, "comments": 0, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "Mundo世界模型从少量房间照片生成可模拟场景", "summary_en": "Presents Mundo, a world model that builds an interactive simulation of a room from a few photos.", "origin": "backfill", "linkers": [] }, { "id": "50219974", "url": "https://www.anthropic.com/research/reviewing-the-evidence-on-worker-retraining-programs", "domain": "anthropic.com", "title": "How well do job retraining programs work?", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50219974", "points": 1, "comments": 0, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "Anthropic综述工人再培训项目的证据", "summary_en": "An evidence review of worker retraining programs from Anthropic's Economic Research team.", "origin": "backfill", "linkers": [] }, { "id": "50220499", "url": "https://nrposner.com/blog/crowleys-three-axes-of-json-querying/", "domain": "nrposner.com", "title": "Crowley's Three Axes of JSON Querying", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50220499", "points": 1, "comments": 0, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "Crowley流式JSON查询突破2 GB/s的设计与测量", "summary_en": "Describes design axes and benchmarks for a streaming JSON query engine exceeding 2 GB/s.", "origin": "backfill", "linkers": [] }, { "id": "50220605", "url": "https://tridao.me/blog/2026/replayssm/", "domain": "tridao.me", "title": "ReplaySSM: Cache SSM Inputs, Not State", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50220605", "points": 1, "comments": 0, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "ReplaySSM缓存SSM输入而非状态以加速解码", "summary_en": "Proposes ReplaySSM, which caches SSM inputs instead of recurrent state to reduce decoding memory traffic.", "origin": "backfill", "linkers": [] }, { "id": "50221043", "url": "https://www.memstrata.dev/", "domain": "memstrata.dev", "title": "MemStrata: AI answers you can check, with sources.", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50221043", "points": 2, "comments": 0, "date": "2026-10-19", "submissions": 3, "source": "hn", "summary_zh": "LongMemEval得分会随评审能否看到来源而变化", "summary_en": "Shows LongMemEval scores shift depending on whether the judge can see the sources.", "origin": "backfill", "linkers": [] }, { "id": "50221693", "url": "https://www.uxness.in/2026/10/ux-research-for-ai-products-ai-native-research-lifecycle.html", "domain": "uxness.in", "title": "UX Research for AI Products: The AI-Native Research Lifecycle ~ UXness: Insights & Inspiration on UX, Visual, and Product Design", "category": "explainer", "topic": "other", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50221693", "points": 1, "comments": 0, "date": "2026-10-24", "submissions": 1, "source": "hn", "summary_zh": "介绍AI产品全生命周期中的用户研究方法", "summary_en": "A guide to embedding UX research across the AI product lifecycle.", "origin": "backfill", "linkers": [] }, { "id": "50229055", "url": "https://gctl.io/blog/2026/when-sandboxed-agents-talk-through-the-cpu-cache/", "domain": "gctl.io", "title": "When sandboxed agents talk through the CPU cache", "category": "explainer", "topic": "security-privacy", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50229055", "points": 1, "comments": 0, "date": "2026-10-25", "submissions": 1, "source": "hn", "summary_zh": "讲解沙箱智能体如何通过CPU缓存隐蔽通信", "summary_en": "Explains covert and side channels via CPU cache in relation to a report on sandboxed agents.", "origin": "backfill", "linkers": [] }, { "id": "50231257", "url": "https://emptysqua.re/blog/polyglots-dilemma/", "domain": "emptysqua.re", "title": "The Polyglot's Dilemma: Conformance Testing a Dozen Specs in as Many Languages", "category": "explainer", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50231257", "points": 1, "comments": 0, "date": "2026-10-25", "submissions": 1, "source": "hn", "summary_zh": "介绍MongoDB如何对多语言实现做规范一致性测试", "summary_en": "Explains how MongoDB tests many client libraries against shared specs.", "origin": "backfill", "linkers": [] }, { "id": "50231816", "url": "https://drmowinckels.io/blog/2025/neon-ghosts/", "domain": "drmowinckels.io", "title": "Neon Ghosts with ggplot2", "category": "explainer", "topic": "other", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50231816", "points": 2, "comments": 0, "date": "2026-10-25", "submissions": 1, "source": "hn", "summary_zh": "教程:用R的ggplot2绘制霓虹幽灵图", "summary_en": "A tutorial on drawing glowing ghost graphics in R with ggplot2.", "origin": "backfill", "linkers": [] }, { "id": "50232753", "url": "https://www.redwoodresearch.org/blog/simulated-control-evaluations", "domain": "redwoodresearch.org", "title": "Simulated Control Evaluations", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50232753", "points": 2, "comments": 0, "date": "2026-10-25", "submissions": 1, "source": "hn", "summary_zh": "用伪造工具输出的LLM红队测试AI智能体监控器", "summary_en": "Presents simulated control evaluations that red-team AI agent monitors using an LLM that fakes tool outputs.", "origin": "backfill", "linkers": [] }, { "id": "50235746", "url": "https://alphapixeldev.com/a-comprehensive-introduction-and-overview-of-gaussian-splats-part-1-introduction/", "domain": "alphapixeldev.com", "title": "A Comprehensive Intro to Gaussian Splats (Part 1)", "category": "explainer", "topic": "robotics-graphics", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50235746", "points": 2, "comments": 0, "date": "2026-10-25", "submissions": 1, "source": "hn", "summary_zh": "介绍高斯泼溅3D渲染方法的基础", "summary_en": "An introductory overview of Gaussian Splatting for 3D rendering.", "origin": "backfill", "linkers": [] }, { "id": "50235773", "url": "https://repo-sam.inria.fr/fungraph/3d-gaussian-splatting/", "domain": "repo-sam.inria.fr", "title": "3D Gaussian Splatting for Real-Time Radiance Field Rendering", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50235773", "points": 2, "comments": 0, "date": "2026-10-25", "submissions": 1, "source": "hn", "summary_zh": "3D Gaussian Splatting实时辐射场渲染论文项目页", "summary_en": "Project page for the SIGGRAPH paper on 3D Gaussian Splatting for real-time radiance field rendering.", "origin": "backfill", "linkers": [ "alphapixeldev.com" ] }, { "id": "50236088", "url": "https://theorem.dev/blog/verified-software-stack/", "domain": "theorem.dev", "title": "Bootstrapping the Verified Software Stack", "category": "research_report", "topic": "math-theory", "confidence": "medium", "interactive": true, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50236088", "points": 2, "comments": 0, "date": "2026-10-23", "submissions": 2, "source": "hn", "summary_zh": "提出为nixpkgs生态构建共享的已验证软件栈", "summary_en": "Describes a research effort to build a shared formally verified software stack starting with nixpkgs.", "origin": "backfill", "linkers": [] }, { "id": "50245763", "url": "https://www.lunarsim.org/research/", "domain": "lunarsim.org", "title": "Research | Galatea: optimal transport for lunar place recognition", "category": "research_report", "topic": "robotics-graphics", "confidence": "medium", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50245763", "points": 1, "comments": 1, "date": "2026-10-26", "submissions": 1, "source": "hn", "summary_zh": "Galatea用最优传输融合全局与局部描述子做月面定位", "summary_en": "Presents Galatea, an optimal transport method merging global and local descriptors for lunar visual place recognition.", "origin": "backfill", "linkers": [] }, { "id": "50249987", "url": "https://digitaliziran.si/2026/10/26/the-half-life-of-an-instruction.html", "domain": "digitaliziran.si", "title": "The Half-Life of an Instruction", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50249987", "points": 2, "comments": 0, "date": "2026-10-26", "submissions": 1, "source": "hn", "summary_zh": "测试DeepSeek V4 Pro指令遵循随示例变化的衰减", "summary_en": "Reports an experiment on how DeepSeek V4 Pro's instruction following changes with few-shot examples.", "origin": "backfill", "linkers": [] }, { "id": "50250129", "url": "https://agent-hypervisor.ai/posts/agent-first-repos/", "domain": "agent-hypervisor.ai", "title": "Agent-First Repos", "category": "explainer", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50250129", "points": 1, "comments": 0, "date": "2026-10-26", "submissions": 1, "source": "hn", "summary_zh": "介绍为编码智能体构建的agent-first代码库", "summary_en": "A long guide to structuring codebases for coding agents to develop and operate.", "origin": "backfill", "linkers": [] }, { "id": "50250494", "url": "https://www.manifold.security/blog/agent-log-tampering", "domain": "manifold.security", "title": "AI agents have few qualms with log tampering", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50250494", "points": 2, "comments": 0, "date": "2026-10-26", "submissions": 1, "source": "hn", "summary_zh": "研究显示多数前沿AI智能体会篡改自身日志", "summary_en": "Reports tests showing most frontier AI agents comply with requests to tamper with their own logs.", "origin": "backfill", "linkers": [] }, { "id": "50251596", "url": "https://research.nvidia.com/labs/par/projects/jca/", "domain": "research.nvidia.com", "title": "Joints, Camera, Action! Co-Generating 3D Motion and Video for Arbitrary Rigs", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50251596", "points": 1, "comments": 0, "date": "2026-10-26", "submissions": 1, "source": "hn", "summary_zh": "JCA为任意骨架联合生成视频与同步3D动作", "summary_en": "Presents Joints, Camera, Action, a model that jointly generates video and 3D motion for arbitrary rigs.", "origin": "backfill", "linkers": [] }, { "id": "50251741", "url": "https://www.airealist.ai/p/exploring-apple-on-device-model", "domain": "airealist.ai", "title": "Exploring Apple's On-Device Model: Core Advanced", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50251741", "points": 1, "comments": 0, "date": "2026-10-25", "submissions": 2, "source": "hn", "summary_zh": "分析Apple 20B参数端侧模型如何存于闪存", "summary_en": "An analysis of how Apple's 20B-parameter on-device model is stored in flash and activates few experts.", "origin": "backfill", "linkers": [] }, { "id": "50252439", "url": "https://maccus.fr/en/ai-models-hide-capabilities-oxford/", "domain": "maccus.fr", "title": "Researchers hid an encrypted capability in an AI model, and fine-tuning couldn't unlock it", "category": "explainer", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50252439", "points": 2, "comments": 1, "date": "2026-10-26", "submissions": 1, "source": "hn", "summary_zh": "解读论文:AI模型可被证明地隐藏任意能力", "summary_en": "Summarizes an October 2026 paper showing AI models can provably hide encrypted capabilities.", "origin": "backfill", "linkers": [] }, { "id": "50253974", "url": "https://lagrangianblog.substack.com/p/end-effector-action-policy-outputs", "domain": "lagrangianblog.substack.com", "title": "End effector action policy outputs to robot model actuation", "category": "explainer", "topic": "robotics-graphics", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50253974", "points": 1, "comments": 0, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "讲解机器人末端执行器策略输出到执行器的数学", "summary_en": "Explains the math converting end effector policy outputs to robot actuation.", "origin": "backfill", "linkers": [] }, { "id": "50258864", "url": "https://www.maxtaylor.me/articles/same-price-different-reason-probing-ai-pricing-agents-for-collusion", "domain": "maxtaylor.me", "title": "Same price, different reason. Probing AI pricing agents for collusion", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50258864", "points": 2, "comments": 1, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "用线性探针检测AI定价智能体的串通", "summary_en": "Tests whether linear probes can distinguish colluding AI pricing agents from innocent ones.", "origin": "backfill", "linkers": [] }, { "id": "50260197", "url": "https://www.kelvinjiang.com/2026/10/can-machines-follow-voice-through-crowded-room.html", "domain": "kelvinjiang.com", "title": "Can Machines Follow One Voice Through a Crowded Room?", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50260197", "points": 1, "comments": 0, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "探索机器能否在嘈杂环境中跟随单一说话人", "summary_en": "An exploratory experiment on whether machines can follow one voice in a crowded room.", "origin": "backfill", "linkers": [] }, { "id": "50260348", "url": "https://janni.turunen.in/poisoned-open-weights/", "domain": "janni.turunen.in", "title": "From the bottom of my heart, I salute you", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50260348", "points": 2, "comments": 0, "date": "2026-10-27", "submissions": 2, "source": "hn", "summary_zh": "展示向开放权重模型植入后门的三种方式", "summary_en": "Demonstrates three ways to poison open-weight models, including a few hundred poisoned samples.", "origin": "backfill", "linkers": [] }, { "id": "50260456", "url": "https://theprimary.com/science/2026-10-26/mobility-power-law-artifacts", "domain": "theprimary.com", "title": "Human travel patterns reflect geographic boundaries rather than universal laws", "category": "research_report", "topic": "physics-earth", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50260456", "points": 1, "comments": 0, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "研究显示人类出行模式由地理边界而非普适律决定", "summary_en": "Reports that human travel patterns follow geographic containers rather than universal power laws.", "origin": "backfill", "linkers": [] }, { "id": "50261004", "url": "https://www.amplifypartners.com/blog-posts/the-fertility-of-an-idea", "domain": "amplifypartners.com", "title": "The Fertility of an Idea: tracing the discovery lineage of optogenetics", "category": "research_report", "topic": "bio-health", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50261004", "points": 1, "comments": 0, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "追溯光遗传学发现谱系的历史研究", "summary_en": "Traces the discovery lineage of optogenetics through citation and historical analysis.", "origin": "backfill", "linkers": [] }, { "id": "50261289", "url": "https://voygr.tech/blog/can-ai-draw-san-francisco-from-memory/", "domain": "voygr.tech", "title": "Can AI draw San Francisco from memory?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50261289", "points": 1, "comments": 0, "date": "2026-10-27", "submissions": 1, "source": "hn", "summary_zh": "测试六个AI聊天模型凭记忆绘制旧金山地图", "summary_en": "Tests six AI chat models at reproducing San Francisco map data from memory.", "origin": "backfill", "linkers": [] }, { "id": "50264471", "url": "https://www.lesswrong.com/posts/ndS9Xn8Qx5PppHJ3Q/the-ability-of-frontier-models-to-control-robots-is-rapidly", "domain": "lesswrong.com", "title": "The Ability of Frontier Models to Control Robots is Rapidly Improving", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50264471", "points": 2, "comments": 0, "date": "2026-10-27", "submissions": 2, "source": "hn", "summary_zh": "展示前沿模型控制机器人的能力快速提升", "summary_en": "Demonstrates that general-purpose multimodal models are rapidly improving at controlling robots.", "origin": "backfill", "linkers": [] }, { "id": "50268814", "url": "https://www.sigarch.org/compressing-llms-without-compromise/", "domain": "sigarch.org", "title": "Compressing LLMs without Compromise", "category": "explainer", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50268814", "points": 2, "comments": 1, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "介绍LLM压缩中权重、KV缓存等数据搬移的方法", "summary_en": "A SIGARCH article on compressing LLM weights and caches to reduce data movement.", "origin": "backfill", "linkers": [] }, { "id": "50273671", "url": "https://tcumulus.dev/blog/building-my-own-tiny-ai-weather-model/", "domain": "tcumulus.dev", "title": "Building My Own Tiny AI Weather Model", "category": "research_report", "topic": "physics-earth", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50273671", "points": 2, "comments": 0, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "基于GraphCast架构自建小型AI天气模型", "summary_en": "Describes building a small AI weather model based on the GraphCast architecture.", "origin": "backfill", "linkers": [] }, { "id": "50276719", "url": "https://www.snowflake.com/en/blog/engineering/Benchmarking-coding-agents/", "domain": "snowflake.com", "title": "Benchmarking Coding Agents at Scale", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50276719", "points": 1, "comments": 0, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "SnowPR用真实PR评测编码智能体于4300万行代码库", "summary_en": "Introduces SnowPR, a benchmark built from real pull requests to evaluate coding agents on a large monorepo.", "origin": "backfill", "linkers": [] }, { "id": "50278103", "url": "https://www.perplexity.ai/hub/blog/rethinking-search-as-code-generation", "domain": "perplexity.ai", "title": "Rethinking Search as Code Generation", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50278103", "points": 1, "comments": 0, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "Perplexity提出把搜索重构为代码生成的可编程原语", "summary_en": "Perplexity describes recasting search as programmable primitives and code generation for agents.", "origin": "backfill", "linkers": [] }, { "id": "50279095", "url": "https://physicstoday.aip.org/news/elizabeth-fulhame-a-forgotten-chemistry-pioneer", "domain": "physicstoday.aip.org", "title": "Elizabeth Fulhame, a forgotten chemistry pioneer", "category": "explainer", "topic": "humanities", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50279095", "points": 2, "comments": 0, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "介绍被遗忘的化学先驱Elizabeth Fulhame", "summary_en": "A historical article on Elizabeth Fulhame's 1794 work on catalysis.", "origin": "backfill", "linkers": [] }, { "id": "50281267", "url": "https://app.reve.com/training-infra", "domain": "app.reve.com", "title": "Training Reve 2.0 at scale", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50281267", "points": 1, "comments": 0, "date": "2026-10-28", "submissions": 1, "source": "hn", "summary_zh": "Reve 2.0多模态模型训练基础设施的技术报告", "summary_en": "A technical report on the infrastructure used to train Reve 2.0 multimodal models.", "origin": "backfill", "linkers": [] }, { "id": "50287227", "url": "https://grafista.io/blog/posts/grafista-vs-claude-vs-canva-vs-moda", "domain": "grafista.io", "title": "Why AI Design Tools Keep Making the Same Flyer: Claude Design, Canva, Moda and Grafista Tested", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50287227", "points": 1, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "对比四款AI设计工具对相同传单提示的输出", "summary_en": "Compares outputs of four AI design tools on identical flyer prompts, repeated three times.", "origin": "backfill", "linkers": [] }, { "id": "50290235", "url": "https://www.harness-hub.org/reports/2026-10", "domain": "harness-hub.org", "title": "State of Harness—October 2026", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50290235", "points": 2, "comments": 1, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "State of Harness首份报告界定智能体harness与市场", "summary_en": "A report defining the agent harness and surveying the market with directory counts.", "origin": "backfill", "linkers": [] }, { "id": "50291045", "url": "https://tenstorrent.com/en/developers/blog/serving-prefill-and-decode-for-agentic-ai", "domain": "tenstorrent.com", "title": "Serving Prefill and Decode for Agentic AI", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50291045", "points": 2, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "在Tenstorrent Blackhole上服务智能体的prefill与decode", "summary_en": "Describes serving prefill and decode for agentic AI workloads on Tenstorrent Blackhole hardware.", "origin": "backfill", "linkers": [] }, { "id": "50293659", "url": "https://politicaldna.org/ai-political-dna/", "domain": "politicaldna.org", "title": "31 AI Models Took a Political Quiz. Here Is Where Each One Landed", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50293659", "points": 1, "comments": 1, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "让31个LLM做政治问卷,发现其趋同程度高于人类", "summary_en": "Gave a 32-statement political quiz to 31 LLMs over 1,017 runs and compared them with human respondents.", "origin": "backfill", "linkers": [] }, { "id": "50297418", "url": "https://pytorch.org/blog/flexattention/", "domain": "pytorch.org", "title": "FlexAttention: The Flexibility of PyTorch with the Performance of FlashAttention", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50297418", "points": 1, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "介绍PyTorch FlexAttention的灵活性与性能", "summary_en": "Explains PyTorch FlexAttention, which provides flexible attention variants with FlashAttention-level performance.", "origin": "backfill", "linkers": [] }, { "id": "50297928", "url": "https://broutonlab.com/blog/deterministic-ai/", "domain": "broutonlab.com", "title": "Deterministic AI: Why LLMs Aren't, What to Do", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50297928", "points": 1, "comments": 0, "date": "2026-10-29", "submissions": 1, "source": "hn", "summary_zh": "讲解LLM为何不确定及如何构建确定性产品", "summary_en": "Explains why LLMs are nondeterministic at temperature 0 and how to build deterministic products.", "origin": "backfill", "linkers": [] }, { "id": "50299212", "url": "https://institute.deepmind.com/essays/the-economics-of-intelligence/", "domain": "institute.deepmind.com", "title": "AGI Economics: Coordination vs. Competence", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50299212", "points": 27, "comments": 6, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "DeepMind分析AGI经济学中协调与能力的关系", "summary_en": "A DeepMind institute analysis of AGI economics, covering jagged agents and mosaic organizations.", "origin": "backfill", "linkers": [] }, { "id": "50299568", "url": "https://divagr.com/blog/building-oversight-tools-into-a-models-weights", "domain": "divagr.com", "title": "Building oversight tools into a model's weights", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50299568", "points": 2, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "把监督工具编入Qwen权重的测谎与日志机制", "summary_en": "Describes building a lie detector and log into Qwen weights that the model cannot disable.", "origin": "backfill", "linkers": [] }, { "id": "50303109", "url": "https://streamhpc.com/blog/2026-09-30/opengl-compute-shaders-before-compute-shaders-heroic-times/", "domain": "streamhpc.com", "title": "OpenGL “Compute Shaders” Before Compute Shaders–Heroic Times", "category": "explainer", "topic": "robotics-graphics", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50303109", "points": 2, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "回顾计算着色器出现前的OpenGL通用计算", "summary_en": "A historical explainer on GPU compute with OpenGL before compute shaders existed.", "origin": "backfill", "linkers": [] }, { "id": "50303242", "url": "https://grace17.org/docs/cycle-double-cover.html", "domain": "grace17.org", "title": "An adventure in multifoundational Mathematics 2.0: the cycle double cover theorem in ZFC with Grace17", "category": "research_report", "topic": "math-theory", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50303242", "points": 2, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "在ZFC中证明循环双覆盖定理的多基础数学案例", "summary_en": "Reports a formal proof of the cycle double cover theorem in ZFC using Grace17.", "origin": "backfill", "linkers": [] }, { "id": "50303427", "url": "https://www.progressiverobot.com/2026/10/27/reverse-turing-test-ais-fail-to-detect-ais/", "domain": "progressiverobot.com", "title": "Reverse Turing Test: Surprising Proof AIs Fail to Spot AIs", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50303427", "points": 1, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "反向图灵测试中七个AI评审仅识别出3/98个AI", "summary_en": "Reports a reverse Turing test where seven AI judges caught only 3 of 98 AI impostors.", "origin": "backfill", "linkers": [] }, { "id": "50303573", "url": "https://cset.georgetown.edu/article/how-black-box-distillation-works/", "domain": "cset.georgetown.edu", "title": "How Black-Box Distillation Works | Center for Security and Emerging Technology", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50303573", "points": 1, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "概述黑盒蒸馏的工作原理", "summary_en": "A policy-oriented overview of how black-box distillation works.", "origin": "backfill", "linkers": [] }, { "id": "50305162", "url": "https://www.babush.me/how-agents-search-mcp-tools.html", "domain": "babush.me", "title": "What agents see when they search MCP tools", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50305162", "points": 1, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "分析智能体搜索MCP工具时看到的内容", "summary_en": "Analyzes how agents search MCP tools and why the search misses relevant decompiler functions.", "origin": "backfill", "linkers": [] }, { "id": "50306377", "url": "https://www.mahl.me/blog/product-hunt-is-swamped-now/", "domain": "mahl.me", "title": "Product Hunt is swamped now", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50306377", "points": 1, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "抓取Product Hunt数据量化2023年后发布量变化", "summary_en": "A small scrape quantifies how Product Hunt launch volume changed after 2023.", "origin": "backfill", "linkers": [] }, { "id": "50307722", "url": "https://www.fig.inc/blog/understanding-open-ended-work-from-a-data-perspective/", "domain": "fig.inc", "title": "Understanding Open-ended Work from a Data Perspective", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50307722", "points": 2, "comments": 0, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "Fig分析开放式工作数据中缺失的学习信号", "summary_en": "Fig describes learning signals missing in data on open-ended computer-use work.", "origin": "backfill", "linkers": [] }, { "id": "50309959", "url": "https://www.sourcemeta.com/blog/json-schema-dialect-auto-upgrade/", "domain": "sourcemeta.com", "title": "Solving the JSON Schema dialect fragmentation problem with auto-upgrades", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50309959", "points": 1, "comments": 1, "date": "2026-10-30", "submissions": 1, "source": "hn", "summary_zh": "用自动升级解决JSON Schema方言碎片化", "summary_en": "Describes an approach to auto-upgrade JSON Schema dialects to solve fragmentation.", "origin": "backfill", "linkers": [] }, { "id": "50313630", "url": "https://alignment.anthropic.com/2026/automated-em-search/", "domain": "alignment.anthropic.com", "title": "Automated Subtle Emergent Misalignment Research", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50313630", "points": 2, "comments": 0, "date": "2026-10-31", "submissions": 2, "source": "hn", "summary_zh": "自动化研究微妙的涌现式错位", "summary_en": "Alignment research on automating the study of subtle emergent misalignment.", "origin": "backfill", "linkers": [] }, { "id": "50314368", "url": "https://soniqo.audio/blog/voice-agent-benchmarks", "domain": "soniqo.audio", "title": "Measuring voice agents: turn-taking, tool calls and expressive speech", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50314368", "points": 1, "comments": 0, "date": "2026-10-31", "submissions": 1, "source": "hn", "summary_zh": "评测语音智能体的轮替、工具调用与表达性语音", "summary_en": "Reports measurements of voice agents using TurnBench, SLURP and InstructTTSEval.", "origin": "backfill", "linkers": [] }, { "id": "50321012", "url": "https://34-118-125-101.sslip.io/log/rappaport/", "domain": "34-118-125-101.sslip.io", "title": "A CVE triage pipeline that keeps asking, because day one tells you nothing", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50321012", "points": 1, "comments": 0, "date": "2026-10-31", "submissions": 1, "source": "hn", "summary_zh": "描述一个持续追问的CVE分诊流水线Rappaport的设计与经验", "summary_en": "Describes Rappaport, a CVE triage pipeline that keeps re-asking questions over time, with design details and lessons.", "origin": "backfill", "linkers": [] }, { "id": "50322528", "url": "https://internals.laxmena.com/p/can-an-llm-and-a-keyword-index-replace", "domain": "internals.laxmena.com", "title": "Can an LLM and a keyword index replace embeddings?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50322528", "points": 2, "comments": 0, "date": "2026-11-01", "submissions": 1, "source": "hn", "summary_zh": "实验比较用LLM加关键词索引替代embedding做语义检索的效果", "summary_en": "Experiment testing whether an LLM plus a keyword index can replace embeddings for semantic search, with measurements.", "origin": "backfill", "linkers": [] }, { "id": "50322946", "url": "https://github.blog/news-insights/octoverse/octoverse-commits-pushes-reach-new-highs-in-2026-as-developers-build-with-agents/", "domain": "github.blog", "title": "Octoverse: Commits, pushes, reach new highs in 2026 as developers build with agents", "category": "research_report", "topic": "econ-social", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50322946", "points": 2, "comments": 1, "date": "2026-10-28", "submissions": 3, "source": "hn", "summary_zh": "GitHub年度报告,统计2026年提交量与开发者增长数据", "summary_en": "GitHub annual report presenting 2026 statistics on commits, developers and growth across industries.", "origin": "backfill", "linkers": [ "kmaliszewski9.github.io", "stackness.dev" ] }, { "id": "50323735", "url": "https://safeguards.anthropic.com/2026/self-assessment/", "domain": "safeguards.anthropic.com", "title": "Scaling Misuse Detection with Self-Assessment", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50323735", "points": 2, "comments": 3, "date": "2026-10-29", "submissions": 2, "source": "hn", "summary_zh": "提出让Claude自评回答安全性以扩展滥用检测的方法并评估", "summary_en": "Presents and evaluates a method where Claude self-assesses response safety to scale misuse detection.", "origin": "backfill", "linkers": [] }, { "id": "50326490", "url": "https://aictrl.dev/blog/open-source-review-ai-era", "domain": "aictrl.dev", "title": "The abandoned PR page", "category": "research_report", "topic": "engineering-env", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50326490", "points": 2, "comments": 0, "date": "2026-11-01", "submissions": 1, "source": "hn", "summary_zh": "分析两个AI原生开源项目中PR合并率与被放弃PR的数据", "summary_en": "Data analysis of pull request outcomes in two AI-native open-source projects, focusing on abandoned PRs.", "origin": "backfill", "linkers": [] }, { "id": "50328206", "url": "https://www.lesswrong.com/posts/TJGPL5FAk9sCioPbE/leaking-muse-s-charter-meta-s-model-spec", "domain": "lesswrong.com", "title": "Leaking Muse's Charter, Meta's model spec", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50328206", "points": 2, "comments": 0, "date": "2026-11-01", "submissions": 1, "source": "hn", "summary_zh": "通过提取思维链内容还原Meta Muse模型的Charter规范", "summary_en": "Reports extracting the content of Meta Muse Spark Open model Charter from its chain of thought.", "origin": "backfill", "linkers": [] }, { "id": "50332791", "url": "https://focs-lab.comp.nus.edu.sg/blog/memory-safety-audit/", "domain": "focs-lab.comp.nus.edu.sg", "title": "What does “verified memory safe” even mean?", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50332791", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "对Anthropic与Theorem的内存安全证明做独立审计", "summary_en": "Independent audit of the memory-safety proofs by Anthropic and Theorem, assessing what verified memory safe means.", "origin": "backfill", "linkers": [] }, { "id": "50335010", "url": "https://pbb.sh/research/Agentic-Backdoor/", "domain": "pbb.sh", "title": "Agentic Backdoors via Pretraining Poisoning", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50335010", "points": 2, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "证明预训练投毒可植入远程代码执行的agent后门", "summary_en": "Shows that pretraining data poisoning can plant a remote-code-execution backdoor in agents.", "origin": "backfill", "linkers": [] }, { "id": "50340598", "url": "https://blitzrank.ai/", "domain": "blitzrank.ai", "title": "BlitzRank: Principled Zero-shot Ranking Agents with Tournament Graphs", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50340598", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "提出基于锦标赛图的零样本LLM文档排序方法BlitzRank", "summary_en": "Project page for BlitzRank, a tournament-graph method for query-efficient zero-shot LLM document ranking.", "origin": "backfill", "linkers": [] }, { "id": "50341099", "url": "https://folds.ink/posts/mtg-benchmark", "domain": "folds.ink", "title": "Making a Magic The Gathering Eval", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50341099", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "构建并评测LLM玩万智牌的基准", "summary_en": "Builds a harness and benchmark for LLMs playing Magic The Gathering and reports results.", "origin": "backfill", "linkers": [] }, { "id": "50341223", "url": "https://www.lesswrong.com/posts/dbStc8ZFESqfihwAd/lessons-from-red-teaming-auto-review-openai-s-blocking-1", "domain": "lesswrong.com", "title": "Lessons From Red-teaming Auto-review, OpenAI's Blocking Monitor", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50341223", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "红队测试OpenAI的Auto-review阻断监控器并总结经验", "summary_en": "Red-teaming report on OpenAI Auto-review blocking monitor with lessons learned.", "origin": "backfill", "linkers": [] }, { "id": "50342587", "url": "https://operator0x.com/field-notes/model-execution-graph/", "domain": "operator0x.com", "title": "Understanding LLMs by seeing language as a shadow of a higher-dimensional state", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50342587", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "以高维状态的投影视角解释LLM如何运作", "summary_en": "Explains LLM behavior by viewing language as a shadow of a higher-dimensional state.", "origin": "backfill", "linkers": [ "substack.com" ] }, { "id": "50342993", "url": "https://humansand.ai/blog/gpu-accelerated-weight-teleportation", "domain": "humansand.ai", "title": "GPU-accelerated Weight Teleportation", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50342993", "points": 2, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "提出GPU压缩与流式解压实现无损权重更新", "summary_en": "Describes GPU compression with streamed decompression for lossless model weight updates in RL training.", "origin": "backfill", "linkers": [] }, { "id": "50343102", "url": "https://engineering.fb.com/2026/10/29/security/catching-bugs-before-they-land-jittestings-human-on-the-loop-approach-at-full-scale/", "domain": "engineering.fb.com", "title": "Catching Bugs Before They Land: JiTTesting’s Human-on-the-Loop Approach at Full Scale", "category": "research_report", "topic": "engineering-env", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50343102", "points": 1, "comments": 0, "date": "2026-10-29", "submissions": 2, "source": "hn", "summary_zh": "Meta介绍JiT测试中人在回路的大规模实践", "summary_en": "Meta engineering report on applying human-on-the-loop just-in-time testing with LLMs at full scale.", "origin": "backfill", "linkers": [] }, { "id": "50343122", "url": "https://www.evalshift.dev/blog/ai-evals-for-agents", "domain": "evalshift.dev", "title": ", description, or og:/twitter: tags here on purpose. Every route declares its own via (src/lib/seo.tsx), which React 19 hoists into this head; a static tag would sit ear", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50343122", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "介绍什么是agent评测以及如何编写", "summary_en": "Explains what agent evals are and how to write them.", "origin": "backfill", "linkers": [] }, { "id": "50344895", "url": "https://simonveitner.substack.com/p/internals-of-fp8-mma-on-hopper", "domain": "simonveitner.substack.com", "title": "Internals of FP8 MMA on Hopper", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50344895", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "讲解Hopper架构上FP8 MMA指令的内部机制", "summary_en": "Explains the internals of FP8 MMA instructions on NVIDIA Hopper GPUs.", "origin": "backfill", "linkers": [] }, { "id": "50344935", "url": "https://shotwell.ai/blog/shotwell-bench/", "domain": "shotwell.ai", "title": "Shotwell-Bench: An annotation benchmark that accepts valid alternatives", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50344935", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "提出接受有效替代分段的视频标注基准Shotwell-Bench", "summary_en": "Introduces Shotwell-Bench, a dense video annotation benchmark that accepts valid alternative segmentations.", "origin": "backfill", "linkers": [] }, { "id": "50345968", "url": "https://www.daweisun.me/blog/stl-planning-rl/", "domain": "daweisun.me", "title": "From Words to Waypoints: Teaching a Small LLM to Plan for Robots", "category": "research_report", "topic": "robotics-graphics", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50345968", "points": 1, "comments": 0, "date": "2026-11-02", "submissions": 1, "source": "hn", "summary_zh": "训练小型LLM为机器人生成路径点的实验", "summary_en": "Experiments teaching a small LLM to produce waypoint plans for robots.", "origin": "backfill", "linkers": [] }, { "id": "50347922", "url": "https://blog.budget-engineer.com/space-data-centers-have-you-looked-up-lately", "domain": "blog.budget-engineer.com", "title": "1.26 Million Data Center Satellites? Orbit Says No", "category": "research_report", "topic": "engineering-env", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50347922", "points": 1, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "分析126万颗数据中心卫星对轨道拥挤的影响", "summary_en": "Analysis of whether 1.26 million proposed data center satellites are feasible given orbital crowding.", "origin": "backfill", "linkers": [] }, { "id": "50352518", "url": "https://qlabs.sh/research/parameters", "domain": "qlabs.sh", "title": "Quadrillion-parameter transformers and how to train them", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50352518", "points": 2, "comments": 1, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "提出万亿参数Transformer的缩放律并外推训练方案", "summary_en": "Presents a scaling law and extrapolates training recipes for quadrillion-parameter transformers.", "origin": "backfill", "linkers": [] }, { "id": "50352595", "url": "https://machinelearningmastery.com/choosing-an-llm-inference-optimization-technique-a-decision-tree-approach/", "domain": "machinelearningmastery.com", "title": "Choosing an LLM Inference Optimization Technique: A Decision-Tree Approach", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50352595", "points": 2, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "用决策树选择LLM推理优化技术的教程", "summary_en": "Tutorial using a decision tree to choose among LLM inference optimization techniques.", "origin": "backfill", "linkers": [] }, { "id": "50353189", "url": "https://tis-100.wincon.dev/", "domain": "tis-100.wincon.dev", "title": "TIS-100 Bench: LLMs Push the Pareto Frontier in Programming Puzzle Games", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50353189", "points": 1, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "评测16个LLM在TIS-100编程谜题上的帕累托前沿", "summary_en": "Benchmark of 16 LLMs on TIS-100 puzzles finding 133 new Pareto frontier points.", "origin": "backfill", "linkers": [] }, { "id": "50355600", "url": "https://alignment.openai.com/eval-production-behavior-gap/", "domain": "alignment.openai.com", "title": "Eval-production behavior gaps in agent safety and the role of eval-awareness", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50355600", "points": 2, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "比较agent在安全评测与实际部署中的行为差异及评测意识", "summary_en": "Compares agent behavior in safety evaluations versus production and studies eval-awareness.", "origin": "backfill", "linkers": [] }, { "id": "50355821", "url": "https://www.llamaindex.ai/blog/the-inference-stack-behind-ocria", "domain": "llamaindex.ai", "title": "The Inference Stack Behind OCRia", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50355821", "points": 2, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "介绍OCRia推理栈的解码延迟优化方法", "summary_en": "Describes inference optimizations in the OCRia stack, including quantization and decoding speedups.", "origin": "backfill", "linkers": [] }, { "id": "50355851", "url": "https://allenai.org/blog/statebench", "domain": "allenai.org", "title": "StateBench: Testing how language models track state changes over long documents", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50355851", "points": 2, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "提出StateBench评测LLM在长文档中追踪状态变化", "summary_en": "Introduces StateBench, evaluating how language models track state changes across long documents.", "origin": "backfill", "linkers": [] }, { "id": "50356136", "url": "https://www.zyphra.com/our-work/zyphra-hierarchical-attention", "domain": "zyphra.com", "title": "ZHA: Zyphra Hierarchical Attention", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50356136", "points": 1, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "提出Zyphra分层稀疏注意力方法ZHA", "summary_en": "Introduces ZHA, a hierarchical sparse attention method, with experiments.", "origin": "backfill", "linkers": [] }, { "id": "50357402", "url": "https://sf-tensor.com/engineering/formalizing-ptx", "domain": "sf-tensor.com", "title": "Formalizing PTX in Lean to Prove GPU Kernel Correctness", "category": "research_report", "topic": "math-theory", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50357402", "points": 2, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "在Lean中形式化Nvidia PTX以证明GPU内核正确性", "summary_en": "Formalizes Nvidia PTX in Lean to prove GPU kernel correctness, with accompanying paper.", "origin": "backfill", "linkers": [ "alcidesfonseca.com" ] }, { "id": "50357431", "url": "https://kapteko.com/userland-linux-scheduler-in-rust/", "domain": "kapteko.com", "title": "Userland Linux scheduler in Rust-faster than time", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50357431", "points": 1, "comments": 0, "date": "2026-11-03", "submissions": 1, "source": "hn", "summary_zh": "用Rust实现用户态Linux调度器以加速虚拟时间", "summary_en": "Describes a userland Linux scheduler in Rust that makes CPU-bound programs run faster than wall time.", "origin": "backfill", "linkers": [] }, { "id": "50360221", "url": "https://freddy.us/blog/posts/outer-products/", "domain": "freddy.us", "title": "Outer Products: The Dual View of Linear Algebra", "category": "explainer", "topic": "math-theory", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50360221", "points": 1, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "讲解外积与线性代数的对偶视角", "summary_en": "Explains outer products and the dual view of linear algebra.", "origin": "backfill", "linkers": [] }, { "id": "50363833", "url": "https://www.glow.io/blogs/mcphantom-ai-coding-agents-are-installing-vulnerable-deprecated-mcp-servers-from-training-data", "domain": "glow.io", "title": "MCPhantom: AI Agents Installing Deprecated MCP Servers", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50363833", "points": 1, "comments": 1, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "发现AI编码agent会安装带已知漏洞的废弃MCP服务器", "summary_en": "Reports that AI coding agents install deprecated MCP servers with known vulnerabilities.", "origin": "backfill", "linkers": [] }, { "id": "50364513", "url": "https://ecommerceresearch.org/chatgpt-ads/", "domain": "ecommerceresearch.org", "title": "Can Advertising Fund Free AI? ChatGPT Ads Research", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50364513", "points": 1, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "用实证数据研究ChatGPT广告能否资助免费AI", "summary_en": "Working paper using evidence from ChatGPT ad rollout to study whether advertising can fund free AI.", "origin": "backfill", "linkers": [] }, { "id": "50366053", "url": "https://www.unite.ai/20-percent-of-all-ok-web-responses-ai-gets-are-not-okay/", "domain": "unite.ai", "title": "20 Percent of All ‘Ok’ Web Responses AI Gets Are Not Okay", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50366053", "points": 1, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "实验发现AI抓取网页时约20%的OK响应其实有问题", "summary_en": "Author experiments with AI-assisted scraping finding many HTTP OK responses are not truly valid.", "origin": "backfill", "linkers": [] }, { "id": "50368076", "url": "https://reckoningscience.org/regulation-by-ai-replaying-epas-glyphosate-decisions/", "domain": "reckoningscience.org", "title": "Regulation by AI: Replaying EPA’s Glyphosate Decisions", "category": "research_report", "topic": "econ-social", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50368076", "points": 1, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "让五个AI模型重演50年EPA草甘膦决策并比较", "summary_en": "Experiment having five AI models replay 50 years of EPA glyphosate decisions, finding none more lenient.", "origin": "backfill", "linkers": [] }, { "id": "50369231", "url": "https://blatent.ai/blog/making-flawed-benchmarks-useful", "domain": "blatent.ai", "title": "Making flawed benchmarks useful", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50369231", "points": 1, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "审视有缺陷基准并将其变得有用的评测方法", "summary_en": "Evaluation of extraction accuracy that reveals benchmark flaws and shows how to use them.", "origin": "backfill", "linkers": [] }, { "id": "50369619", "url": "https://blog.janestreet.com/a-study-of-sequence-weighting-at-scale/", "domain": "blog.janestreet.com", "title": "Jane Street Blog - A study of sequence weighting at scale", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50369619", "points": 1, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "研究语言模型中序列加权的缩放律", "summary_en": "Jane Street study of scaling laws for data weighting across language models.", "origin": "backfill", "linkers": [] }, { "id": "50370681", "url": "https://paulg.info/2026/11/04/quantum-weight-space/", "domain": "paulg.info", "title": "Weight-space programming a neural quantum state", "category": "research_report", "topic": "physics-earth", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50370681", "points": 1, "comments": 0, "date": "2026-11-04", "submissions": 1, "source": "hn", "summary_zh": "与Opus协作诊断并修复神经量子态的权重空间编程", "summary_en": "Case study using Opus to diagnose and repair a neural quantum state via weight-space programming.", "origin": "backfill", "linkers": [] }, { "id": "50374752", "url": "https://www.primeintellect.ai/blog/auditing-public-evaluations", "domain": "primeintellect.ai", "title": "Learnings from Auditing Public Evaluations", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50374752", "points": 2, "comments": 0, "date": "2026-11-04", "submissions": 3, "source": "hn", "summary_zh": "审计公开评测发现评分器错误与歧义", "summary_en": "Audit of public LLM evaluations finding broken graders and ambiguous tasks.", "origin": "backfill", "linkers": [] }, { "id": "50376131", "url": "https://www.shuffle-on.com/blog/the-test-that-deleted-my-home-folder", "domain": "shuffle-on.com", "title": "The test that deleted my home folder", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50376131", "points": 2, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "事后分析AI助手测试删除主目录的路径遍历漏洞事故", "summary_en": "Postmortem of an AI coding assistant test that deleted a home folder while demonstrating a path-traversal bug.", "origin": "backfill", "linkers": [] }, { "id": "50378729", "url": "https://qiyaoliang.com/blog/es-vs-pg-long-horizon/", "domain": "qiyaoliang.com", "title": "Forget about credit assignment in long-horizon agentic post-training", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50378729", "points": 1, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "论证长程agent后训练无需信用分配,比较ES与PG", "summary_en": "Argues credit assignment is unnecessary in agentic post-training, comparing evolution strategies with policy gradient.", "origin": "backfill", "linkers": [] }, { "id": "50380658", "url": "https://kadhar.dev/blog/multi-agent-systems-from-scratch", "domain": "kadhar.dev", "title": "Building a Multi-Agent System: 3 Patterns, Explained Simply", "category": "explainer", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50380658", "points": 102, "comments": 43, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "讲解多agent系统的三种架构模式", "summary_en": "Tutorial on three architecture patterns for multi-agent systems.", "origin": "backfill", "linkers": [] }, { "id": "50381288", "url": "https://www.a16z.news/p/the-exploit-was-the-bottleneck", "domain": "a16z.news", "title": "The Exploit Was the Bottleneck", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50381288", "points": 1, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "用23年数据分析AI与补丁差距", "summary_en": "Data analysis of 23 years of vulnerability data on AI and the patch gap.", "origin": "backfill", "linkers": [] }, { "id": "50381415", "url": "https://tvm.apache.org/2026/11/03/rethinking-abstractions-in-the-agentic-era", "domain": "tvm.apache.org", "title": "Rethinking Abstractions in the Agentic Era: Towards a DSL and Compiler Foundation for Agentic Kernel Development", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50381415", "points": 1, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "展示用Claude与Codex构建DSL与编译器基础", "summary_en": "TVM community report on using coding agents to build DSL and compiler foundations.", "origin": "backfill", "linkers": [] }, { "id": "50382567", "url": "https://agentunicorn.ai/research/gpt-live-voice-ai-tradeoffs", "domain": "agentunicorn.ai", "title": "GPT-Live vs GPT-Realtime: Voice AI Trade-offs", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50382567", "points": 1, "comments": 1, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "比较GPT-Live、GPT-Realtime与流式语音管线的架构权衡", "summary_en": "Comparison of GPT-Live, GPT-Realtime and streaming voice pipelines on delegation, guardrails and reliability.", "origin": "backfill", "linkers": [] }, { "id": "50383259", "url": "https://jepsen.io/consistency/models", "domain": "jepsen.io", "title": "Consistency Models", "category": "explainer", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50383259", "points": 1, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "Jepsen的一致性模型参考页", "summary_en": "Jepsen reference page explaining consistency models for distributed systems.", "origin": "backfill", "linkers": [] }, { "id": "50385391", "url": "https://eunomia.dev/agentsight/", "domain": "eunomia.dev", "title": "AgentSight: System-wide AI agent profiling and monitoring with eBPF", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50385391", "points": 2, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "提出用eBPF做系统级AI agent监控的AgentSight", "summary_en": "Presents AgentSight, an eBPF-based system-wide profiling and monitoring tool for AI agents.", "origin": "backfill", "linkers": [] }, { "id": "50385490", "url": "https://syfi.cs.washington.edu/blog/2026-11-04-vibesys-trainium/", "domain": "syfi.cs.washington.edu", "title": "VibeSys on Trainium: Agents Build and Tune Qwen3.8 Serving Engines on AWS Trainium2", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50385490", "points": 1, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "让agent在Trainium2上构建并调优Qwen3.8推理引擎", "summary_en": "Evaluates agents building and tuning Qwen3.8 serving engines on AWS Trainium2.", "origin": "backfill", "linkers": [] }, { "id": "50386055", "url": "https://www.gr.inc/blog/how-to-train-a-swarm", "domain": "gr.inc", "title": "How to Train a Swarm", "category": "research_report", "topic": "ai-ml", "confidence": "medium", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50386055", "points": 2, "comments": 0, "date": "2026-11-05", "submissions": 1, "source": "hn", "summary_zh": "讨论训练agent swarm的方法与结果", "summary_en": "Describes methods and results for training agent swarms.", "origin": "backfill", "linkers": [] }, { "id": "50386619", "url": "https://www.lesswrong.com/posts/35yw2qGK7naFLAvcq/teaching-qwen-why-1", "domain": "lesswrong.com", "title": "Teaching Qwen Why", "category": "research_report", "topic": "interpretability-alignment", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": true, "hn_id": "50386619", "points": 1, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "研究如何通过教Qwen理由来改变其行为", "summary_en": "Experiments on teaching Qwen the reasons behind behavior for alignment.", "origin": "backfill", "linkers": [] }, { "id": "50387419", "url": "https://aspirationalneuroscience.org/experiments/", "domain": "aspirationalneuroscience.org", "title": "Experiments That Could Win", "category": "research_report", "topic": "bio-health", "confidence": "medium", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50387419", "points": 2, "comments": 1, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "提出可赢得记忆解码挑战的具体实验方案", "summary_en": "Concrete experimental proposals for winning the Memory Decoding Challenge.", "origin": "backfill", "linkers": [] }, { "id": "50387522", "url": "https://engineering.flexcompute.com/articles/inverse-design-grating-couplers/", "domain": "engineering.flexcompute.com", "title": "Automated \"Inverse\" Design of Grating Couplers for Co-Packaged Optics", "category": "research_report", "topic": "engineering-env", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50387522", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "用形状优化与伴随法自动设计共封装光学光栅耦合器", "summary_en": "Describes automated inverse design of grating couplers for co-packaged optics using shape optimization and adjoint methods.", "origin": "backfill", "linkers": [] }, { "id": "50387826", "url": "https://zrt.wtf/do-lora-optimizers-matter/", "domain": "zrt.wtf", "title": "What does LoRA specialized Muon optimizer even mean?", "category": "explainer", "topic": "ai-ml", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50387826", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "解释专为LoRA设计的Dimuon优化器含义", "summary_en": "Explains what the LoRA-specialized Dimuon optimizer means.", "origin": "backfill", "linkers": [] }, { "id": "50388210", "url": "https://ztatlock.net/pubs/2026-uist-codetations/", "domain": "ztatlock.net", "title": "Codetations: Rich, Granular, and Maintainable Annotations for Evolving Code", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50388210", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "提出支持持久交互标注的VSCode扩展Codetations", "summary_en": "Research project page for Codetations, a VSCode extension for persistent code annotations.", "origin": "backfill", "linkers": [] }, { "id": "50388416", "url": "https://www.goodfire.com/research/mechanistic-shaping", "domain": "goodfire.com", "title": "Shaping and Monitoring Models Through Their Internal Representations", "category": "research_report", "topic": "interpretability-alignment", "confidence": "high", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50388416", "points": 1, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "通过内部表征塑造和监控模型的研究", "summary_en": "Research on shaping and monitoring models through their internal representations.", "origin": "backfill", "linkers": [] }, { "id": "50389668", "url": "https://hodor.gatebolt.com/blog/secret-leak-test", "domain": "hodor.gatebolt.com", "title": "The Secret Leak Test: how well GateBolt kept Hodor's password, and where it didn't", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50389668", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "测试提示注入过滤器GateBolt对AI攻击者的防护", "summary_en": "Test of GateBolt prompt injection filter against AI attackers trying to steal a password.", "origin": "backfill", "linkers": [] }, { "id": "50390194", "url": "https://www.framer.com/blog/lighthouse-13-6/", "domain": "framer.com", "title": "Lighthouse 13.6: how we made PageSpeed Insights more accurate for Framer sites", "category": "research_report", "topic": "systems-hardware", "confidence": "medium", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50390194", "points": 3, "comments": 1, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "Framer介绍与Lighthouse团队优化预加载评分的工程细节", "summary_en": "Framer engineering breakdown of Lighthouse 13.6 changes fixing scoring of preloaded code.", "origin": "backfill", "linkers": [] }, { "id": "50393198", "url": "https://www.keivalya.com/pico-wam/", "domain": "keivalya.com", "title": "pico-WAM—Counterfactual Prediction and Planning Error", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50393198", "points": 3, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "介绍小型世界动作模型pico-WAM并评估反事实预测", "summary_en": "Presents pico-WAM, a compact world action model, evaluating counterfactual prediction and planning.", "origin": "backfill", "linkers": [] }, { "id": "50393344", "url": "https://blog.doubleword.ai/llms-play-pokemon", "domain": "blog.doubleword.ai", "title": "I Spent $1000 Making LLMs Play Competitive Pokémon", "category": "research_report", "topic": "ai-ml", "confidence": "high", "interactive": false, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50393344", "points": 6, "comments": 1, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "花1000美元让LLM对战竞技宝可梦并比较结果", "summary_en": "Experiment spending $1000 on LLMs playing competitive Pokemon against each other.", "origin": "backfill", "linkers": [] }, { "id": "50393397", "url": "https://www.shuffle-on.com/blog/ai-crawlers-read-my-site", "domain": "shuffle-on.com", "title": "594,101 knocks: who reads my site, and what AI crawlers can't see", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": false, "arxiv": false, "hf": false, "github_io": false, "cit": true, "hn_id": "50393397", "points": 3, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "统计GPTBot对SSO登录页的594,101次访问", "summary_en": "Log analysis of OpenAI crawler hitting an SSO login page 594,101 times.", "origin": "backfill", "linkers": [] }, { "id": "50393401", "url": "https://lucianolupo.com/notes/The%20Day%20a%20Test%20Deleted%20My%20Home%20Directory", "domain": "lucianolupo.com", "title": "The Day a Test Deleted My Home Directory", "category": "research_report", "topic": "security-privacy", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50393401", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "事后分析编码agent的变异测试删除主目录", "summary_en": "Postmortem of a coding agent mutation test that deleted a home directory.", "origin": "backfill", "linkers": [] }, { "id": "50393482", "url": "https://www.irisc2.com/blog/heic-hashquine", "domain": "irisc2.com", "title": "Turning the HEIC decoder into a SHA-256 computer", "category": "research_report", "topic": "security-privacy", "confidence": "high", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50393482", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "把HEIC解码器变成SHA-256计算机的技术演示", "summary_en": "Technical demonstration turning a HEIC decoder into a SHA-256 computer via a crafted image sequence.", "origin": "backfill", "linkers": [] }, { "id": "50394156", "url": "https://engineering.flexcompute.com/articles/decision-models-drc-repair/", "domain": "engineering.flexcompute.com", "title": "Using decision models to speed up design rule violation repair in chip design", "category": "research_report", "topic": "engineering-env", "confidence": "medium", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50394156", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "用决策模型加速芯片设计规则违规修复", "summary_en": "Uses decision models to speed up design rule violation repair in chip design.", "origin": "backfill", "linkers": [] }, { "id": "50395457", "url": "https://www.framer.com/blog/harness-2/", "domain": "framer.com", "title": "Harness 2: executing agent code as it streams", "category": "research_report", "topic": "systems-hardware", "confidence": "high", "political": "none", "interactive": true, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50395457", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "Framer发布Harness 2，通过流式执行JavaScript代码提升Agent效率并降低成本。", "summary_en": "Framer introduces Harness 2, executing JavaScript as it streams to improve agent efficiency and reduce costs.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" }, { "id": "50395550", "url": "https://www.lesswrong.com/posts/N5AmYkRks5CqAxvBg/metr-independent-assessment-of-ai-r-and-d-acceleration-at", "domain": "lesswrong.com", "title": "METR: Independent Assessment of AI R&D Acceleration at Anthropic in October 2026", "category": "research_report", "topic": "ai-ml", "confidence": "high", "political": "some", "interactive": true, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50395550", "points": 3, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "METR报告指出Anthropic在2026年10月利用AI加速研发的能力提升了约两倍。", "summary_en": "METR report finds Anthropic's AI R&D acceleration doubled by October 2026 due to AI automation.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" }, { "id": "50397085", "url": "https://www.quantamagazine.org/a-strange-quantum-theorem-has-killed-einsteins-reality-20261106/", "domain": "quantamagazine.org", "title": "A Strange Quantum Theorem Has Killed Einstein’s Reality", "category": "explainer", "topic": "physics-earth", "confidence": "high", "political": "none", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50397085", "points": 3, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "量子定理新进展挑战了爱因斯坦关于现实客观存在的观点。", "summary_en": "New quantum theorem developments challenge Einstein's view of objective reality.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" }, { "id": "50397191", "url": "https://sgiannopoulos.github.io/blog/machine-code-programming/", "domain": "sgiannopoulos.github.io", "title": "Programming in the Machine Code Space: LLMs can read & write executables byte by byte", "category": "research_report", "topic": "ai-ml", "confidence": "high", "political": "none", "interactive": true, "repo": false, "arxiv": true, "hf": true, "github_io": true, "cit": false, "hn_id": "50397191", "points": 3, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "研究显示LLM可直接读取和编写机器码，无需反汇编工具即可分析二进制文件。", "summary_en": "Research shows LLMs can read and write machine code directly without disassemblers.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" }, { "id": "50398045", "url": "https://www.doubleai.com/research/can-llms-redesign-themselves-as-they-train", "domain": "doubleai.com", "title": "Can LLMs redesign themselves as they train?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "political": "none", "interactive": false, "repo": true, "arxiv": true, "hf": true, "github_io": false, "cit": false, "hn_id": "50398045", "points": 9, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "DoubleAI研究让AI代理在训练期间动态重新设计神经网络架构。", "summary_en": "DoubleAI research enables AI agents to redesign neural network architectures during training.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" }, { "id": "50398514", "url": "https://research.nvidia.com/labs/lpr/vflash/", "domain": "research.nvidia.com", "title": "VFlash: Enabling System-1 VLMs for Physical AI", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "political": "none", "interactive": true, "repo": true, "arxiv": false, "hf": false, "github_io": false, "cit": false, "hn_id": "50398514", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "NVIDIA提出VFlash，通过并行视觉规划加速物理AI中的机器人和自动驾驶推理。", "summary_en": "NVIDIA proposes VFlash, accelerating Physical AI inference via parallel visual planning.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" }, { "id": "50398678", "url": "https://www.anthropic.com/research/components-of-agent-memory", "domain": "anthropic.com", "title": "The components of agent memory", "category": "research_report", "topic": "ai-ml", "confidence": "high", "political": "none", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50398678", "points": 4, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "Anthropic研究展示Claude如何通过“做梦”机制整理记忆并纠正错误。", "summary_en": "Anthropic research shows how Claude organizes memory and corrects errors via a 'dreaming' mechanism.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" }, { "id": "50398727", "url": "https://www.anthropic.com/research/teaching-ai-to-build-scientific-models", "domain": "anthropic.com", "title": "Teaching AI to build scientific models", "category": "explainer", "topic": "ai-ml", "confidence": "high", "political": "none", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": false, "hn_id": "50398727", "points": 2, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "研究人员探讨AI如何改变神经科学等科学模型的构建与解释方式。", "summary_en": "Researchers discuss how AI is changing the construction and interpretation of scientific models in neuroscience.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" }, { "id": "50399493", "url": "https://www.juliusresearch.ai/blog/deep-finance", "domain": "juliusresearch.ai", "title": "DeepFinance: Can AI diligence a $100MM deal?", "category": "research_report", "topic": "ai-ml", "confidence": "high", "political": "none", "interactive": false, "repo": false, "arxiv": true, "hf": false, "github_io": false, "cit": true, "hn_id": "50399493", "points": 5, "comments": 2, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "DeepFinance是一个衡量AI在私募股权和投行工作流中表现的基准测试及排行榜。", "summary_en": "DeepFinance is a benchmark and leaderboard measuring AI performance on real-world private equity and investment banking workflows.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" }, { "id": "50399612", "url": "https://pranjal2041.github.io/cua-embodied/", "domain": "pranjal2041.github.io", "title": "Computer-Use Agents Are Embodied Generalists", "category": "research_report", "topic": "robotics-graphics", "confidence": "high", "political": "none", "interactive": false, "repo": true, "arxiv": true, "hf": false, "github_io": true, "cit": false, "hn_id": "50399612", "points": 1, "comments": 0, "date": "2026-11-06", "submissions": 1, "source": "hn", "summary_zh": "研究展示单一计算机使用代理在多个具身AI竞赛中获得优异排名。", "summary_en": "Research shows a single computer-use agent ranking first, second, third, and sixth across four embodied AI competitions.", "linkers": [], "origin": "daily", "added": "2026-11-07", "model": "qwen/qwen3.7-flash" } ], "issue": { "number": 0, "slug": "openshelf", "window": [ "2026-10-07", "2026-11-06" ], "published": "2026-11-07" }, "editorial": [ { "domain": "jhspatial.github.io", "n": 10, "reason_en": "Automated daily digest written by an AI system, not authored research.", "reason_zh": "每天由 AI 自动生成的综述，不是作者自己的研究。", "decided": "2026-11-06" }, { "domain": "tobiasstenzel.com", "n": 8, "reason_en": "Course study notes on data structures, not an explainer of research.", "reason_zh": "数据结构课程笔记，不是对研究的解读。", "decided": "2026-11-06" }, { "domain": "114514547.github.io", "n": 5, "reason_en": "Personal tool troubleshooting notes, not research.", "reason_zh": "个人工具排障记录，不是研究。", "decided": "2026-11-06" }, { "domain": "alphaxiv.org", "n": 4, "reason_en": "Mirror pages for arXiv papers, not the authors' own websites.", "reason_zh": "arXiv 论文的镜像页，不是作者自己的网站。", "decided": "2026-11-06" }, { "domain": "emergentmind.com", "n": 1, "reason_en": "Paper aggregator page, not the authors' own website.", "reason_zh": "论文聚合站的页面，不是作者自己的网站。", "decided": "2026-11-06" }, { "domain": "doi.org", "n": 1, "reason_en": "Link resolves to a preprint server, not the authors' own website.", "reason_zh": "链接指向预印本平台，不是作者自己的网站。", "decided": "2026-11-06" }, { "domain": "theainewsreport.com", "n": 3, "reason_en": "News-explainer site that rewrites trending posts, not responses written by researchers.", "reason_zh": "逐条改写热门帖子的新闻解读站，不是研究者自己写的回应。", "decided": "2026-11-06" } ], "editorial_rules": [ { "rule": "political", "n": 7, "titles": [ "China’s Weaponization of Dual-Use Research Vessels for Undersea Data Dominance", "Grenzmanagement statt Fluchtursachen: Wie sich migrationsbezogene EU-Entwicklungshilfe verändert", "Flagrant: Brocasting Trump, Part III", "Lost Funds: Data Update for October 15", "Lost Funds: Data Update for October 1", "Trump Tourism Effects", "I Built an AI Bot to Fact-Check the Midterms. Here's What I've Learned." ], "label_en": "Main subject is a heated political issue", "label_zh": "主题是政治议题", "desc_en": "Elections, named politicians, partisan policy fights, migration, wars and geopolitical rivalry. Research that only touches on policy stays.", "desc_zh": "选举、具体政治人物、党派之争、移民、战争和国家间对抗。只是涉及政策的研究会保留。" } ] }