37:I[923811,["/_next/static/chunks/1ofa4sofdjo4d.js","/_next/static/chunks/3-z86nrqbt-gd.js","/_next/static/chunks/0sjrm73dthvxo.js","/_next/static/chunks/1apkigbabp_jz.js","/_next/static/chunks/1poz9bl8er6mw.js","/_next/static/chunks/1vmlkhm1uomg0.js","/_next/static/chunks/3b4fzrh8g4nh5.js","/_next/static/chunks/3x8n7joqz45xz.js","/_next/static/chunks/1xfxwjxbn4vd9.js","/_next/static/chunks/0717u_rli2lmr.js","/_next/static/chunks/0oq47q7xu8p9i.js","/_next/static/chunks/3rglvh2uekf89.js","/_next/static/chunks/26mjsq_ce17ei.js","/_next/static/chunks/3q8rs7ais2_qk.js","/_next/static/chunks/1nczxl13y2mf0.js","/_next/static/chunks/2egt9jyrzvvz5.js"],"HomeFeed"] 38:T543,This week, the Gates Foundation announced a $1 billion commitment to build and deliver equitable AI tools to schools, frontline health care workers, farmers, and more. Of that commitment, 40% will go toward education, including AI tutoring to individualize learning in classrooms both in the U.S. and abroad. That’s a whopping $400 million infusion into schools for a tool that’s only been around (at least commercially) for a handful of years. The AI push is part of a broader U.S. education strategy the foundation laid out in July, which aims to help up to 10 million more Americans earn “credentials of value” by 2045, the year the foundation plans to close. And it comes at a time when the value of classroom technology is being questioned. Test scores have dropped among some students over the same time period since computer assessments were widely implemented. Meanwhile, research shows that ever since the U.S. spent $30 billion to swap textbooks for laptops, Gen Z has shown to be less cognitively capable than previous generations, according to written testimony by neuroscientist Jared Cooney Horvath presented to the U.S. Senate Committee on Commerce earlier this year. Read more [paywall removed for Redditors]: https://fortune.com/2026/09/19/the-gates-foundation-ai-schools-teachers-warn-gap-reading-math/?utm_source=reddit/39:T7d4,Hi everyone! GitHub Copilot for JetBrains v1.18.0 brings side-by-side session view, Assisted Approvals (Public Preview), and the ability to re-edit earlier prompts. We’ve also added organization and enterprise customizations, finer MCP tool controls, and Plan mode for Codex. Highlights and fixes below. Full release notes and download . New Features Assisted Approvals (Public Preview): Let AI approve low-risk tool calls in Copilot agent sessions, with approval prompts for higher-risk actions. Side-by-side session view: Chat in the editor area while browsing sessions in the session window (Entry point is at the session toolbar) Re-edit previous messages: Revise an earlier prompt in a Copilot agent session, with the conversation and file changes rewound before the replacement message is sent. Built-in GitHub MCP server toggle: Turn the built-in server on or off through a dedicated setting, enabled by default. Manually configured MCP servers are unaffected. Organization-managed skills: Use shared skills in Local and Copilot agent sessions. Organization-managed custom instructions: Apply your organization’s instructions in Local and Copilot agent sessions. Plan mode for Codex: Review, refine, or approve a plan before the agent starts implementation. Persistent MCP tool controls: Keep your per-tool selections for MCP servers used by Copilot agent sessions. Usage Tips: Browse tips above the chat input, with shortcuts to commands, customizations, and settings. UX Improvements Simplified the chat welcome screen and added a direct Share feedback link. Clearly identified the built-in GitHub MCP server in Configure Tools , with a shortcut to its settings. Bug Fixes Fixed Inline Chat Ask overwriting your edits when a request finished or the chat was closed. Fixed Inline Chat ignoring the selected Thinking Effort and Context Window settings in Copilot agent sessions. Resolved errors and unresponsive controls when stopping or closing Inline Chat, including during session setup. Fi3a:T7de,Hey everyone, I’m sharing a research project I’ve been working on called WarpState — an experimental language-model architecture exploring whether we can move away from storing an ever-growing history of token-level KV states and instead maintain a compact recurrent memory. The basic idea is: exact causal self-attention inside small, non-overlapping token chunks; two associative matrix memories per attention head; a fast memory and a slow memory , with separately learned decay rates; bounded key/value writes from completed chunks; an input-dependent gate that decides how much to use local attention vs. recurrent memory; optional parameter sharing across logical depth. The current reference preset is about 2.005B parameters , with: 28 logical layers / 28 physical cores width 2048 32 attention heads 128-token chunks FFN width 7168 vocabulary size 32,768 https://preview.redd.it/pa7xlbp2hnqh1.png?width=1280&format=png&auto=webp&s=14e87c7f750dfab0133ab92513680fd5b110d357 https://preview.redd.it/dw29eok3hnqh1.png?width=964&format=png&auto=webp&s=dd54e98df212f9245572012715b733c14fba65dd https://preview.redd.it/qup0dol4hnqh1.png?width=1280&format=png&auto=webp&s=35b44a293934d6d1ca78212c8e6cd942aa5a0286 Why two memories? Instead of keeping one recurrent state, WarpState keeps two exponentially decaying associative memories. At initialization, their decay values are approximately: Fast: β ≈ 0.90 Slow: β ≈ 0.99 With 128-token chunks, their ideal isolated-contribution half-lives are roughly: Fast: ~842 tokens Slow: ~8,828 tokens The model learns the decay coefficients, and every head also learns how to mix the two memories. This does not mean that the model has an 8K semantic memory limit — or that information is guaranteed to survive for that long. Those values only describe the initial exponential weighting of isolated writes. Actual usable retention depends on learned decays, interference, queries, training, and how information propagates through depth. Fixed-size streaming 35:["$","$L37",null,{"filterSourceIds":null,"filterFavorites":false,"filterAdminPicks":false,"initialViewMode":"list","initialFeed":{"key":"11|latest||0|0||","rows":[{"id":8862010,"title":"Amazon launches research initiative with Stanford University to advance AI and science","titleKo":"Amazon Launches AI, Energy, and Healthcare Research Initiative with Stanford","hookTitleKo":null,"thumbnailUrl":null,"url":"https://www.amazon.science/news/amazon-launches-research-initiative-with-stanford-university-to-advance-ai-and-science","canonicalUrl":"https://www.amazon.science/news/amazon-launches-research-initiative-with-stanford-university-to-advance-ai-and-science","description":"The collaboration aims to advance research while broadening participation and translating discovery into real-world solutions.","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T19:31:59+00:00","publishedToSiteAt":"2026-09-21T20:07:32.034643+00:00","createdAt":"2026-09-21T20:04:34.80487+00:00","sourceId":7,"isAdminPick":false,"sourceName":"Amazon Science","sourceSlug":"amazon-science","sourceIconUrl":"https://www.amazon.science/apple-touch-icon.png","sourceUrl":"https://www.amazon.science","sourceTier":1},{"id":8920672,"title":"The Business of Building God (13 minute read)","titleKo":"OpenAI and Anthropic's RSI Strategy and the Uncertainty of AI Profitability","hookTitleKo":null,"thumbnailUrl":null,"url":"https://www.strangeloopcanon.com/p/the-business-of-building-god?utm_source=tldrai","canonicalUrl":"https://www.strangeloopcanon.com/p/the-business-of-building-god","description":"AI labs like OpenAI and Anthropic lead with a small model advantage but face competition as costs rise and open-source models close the gap. To sustain profits, labs aim to expand into new sectors, ranging from ads to robotics, while also exploring automation through recursive self-improvement (RSI). However, the path is fraught with uncertainty as enormous financial and talent investments may only maintain a temporary edge amid shifting economic and regulatory landscapes.","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T19:07:03+00:00","publishedToSiteAt":"2026-09-22T17:03:48.442398+00:00","createdAt":"2026-09-22T16:55:56.65085+00:00","sourceId":14,"isAdminPick":false,"sourceName":"TLDR AI","sourceSlug":"tldr-ai","sourceIconUrl":"https://tldr.tech/apple-touch-icon.png","sourceUrl":"https://tldr.tech/ai","sourceTier":2},{"id":8920669,"title":"Meta's Muse personal AI agent tops ChatGPT, Grok and Claude for post-launch downloads (4 minute read)","titleKo":"Meta's 'Muse' Surpasses 2.5 Million Downloads Two Weeks After Launch, Overtaking ChatGPT on iOS","hookTitleKo":null,"thumbnailUrl":"https://image.cnbcfm.com/api/v1/image/108365773-17900129531790012949-48520931775-1080pnbcnews.jpg?v=1790012951&w=750&h=422&vtcrop=y","url":"https://www.cnbc.com/2026/09/21/meta-muse-personal-ai-agent-downloads.html?utm_source=tldrai","canonicalUrl":"https://www.cnbc.com/2026/09/21/meta-muse-personal-ai-agent-downloads.html","description":"Meta's Muse AI app has surged to the top of the iOS App Store charts in the US, outpacing ChatGPT and other AI tools with 730,000 downloads shortly after its release. Powered by the Muse Spark AI models, the app allows users to manage digital assistants for tasks like filling forms and organizing emails. While Muse's popularity highlights growing consumer interest, privacy concerns persist, with Amazon blocking the app over security risks, although Shopify has partnered with Meta to integrate agentic checkout.","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T18:28:51+00:00","publishedToSiteAt":"2026-09-22T17:04:04.883708+00:00","createdAt":"2026-09-22T16:55:56.65085+00:00","sourceId":14,"isAdminPick":false,"sourceName":"TLDR AI","sourceSlug":"tldr-ai","sourceIconUrl":"https://tldr.tech/apple-touch-icon.png","sourceUrl":"https://tldr.tech/ai","sourceTier":2},{"id":8857434,"title":"NVIDIA Launches DSX Ready to Qualify Power and Cooling Products for AI Factories","titleKo":"NVIDIA Unveils 'DSX Ready' Certification Program for Power and Cooling Products for AI Factories","hookTitleKo":null,"thumbnailUrl":"https://blogs.nvidia.com/wp-content/uploads/2026/09/end-to-end-press-dsx-ready-kv-1920x1080-1.png","url":"https://blogs.nvidia.com/blog/dsx-ready-ai-factories-power-cooling/","canonicalUrl":"https://blogs.nvidia.com/blog/dsx-ready-ai-factories-power-cooling","description":"Every AI factory needs power and cooling that fit its computing architecture. As AI infrastructure expands, power, cooling, water, site and grid constraints are shaping what builders can deploy. Choosing products that fit the complete factory design helps builders turn computing capacity into useful AI output. To help builders make those decisions, NVIDIA is introducing […]","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T18:00:07+00:00","publishedToSiteAt":"2026-09-21T18:10:45.838939+00:00","createdAt":"2026-09-21T18:07:59.765305+00:00","sourceId":8,"isAdminPick":false,"sourceName":"NVIDIA Blog","sourceSlug":"nvidia-blog","sourceIconUrl":null,"sourceUrl":"https://blogs.nvidia.com","sourceTier":1},{"id":8856634,"title":"Advancing AI for biology: Teaching models to design and characterize antibodies","titleKo":"Amazon Releases Three Papers on AI-Based Antibody Design and Evaluation","hookTitleKo":null,"thumbnailUrl":"https://cdn.amazon.science/dims4/default/cf81c8b/2147483647/strip/true/crop/1920x1008+0+36/resize/1200x630!/quality/90/?url=https%3A%2F%2Famzn-science-production-science.s3.us-east-1.amazonaws.com%2Fscience%2Ffd%2F70%2F3a65a679455aba2bf7e351dff7aa%2Fbiodiscovery-16x9.png","url":"https://www.amazon.science/blog/advancing-ai-for-biology-teaching-models-to-design-and-characterize-antibodies","canonicalUrl":"https://www.amazon.science/blog/advancing-ai-for-biology-teaching-models-to-design-and-characterize-antibodies","description":"Three new papers from Amazon Bio Discovery address bottlenecks in AI-driven antibody engineering, from benchmarking binding predictors to experimentally validating de novo design.","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T17:51:00+00:00","publishedToSiteAt":"2026-09-21T18:10:00.264623+00:00","createdAt":"2026-09-21T18:05:32.204673+00:00","sourceId":7,"isAdminPick":false,"sourceName":"Amazon Science","sourceSlug":"amazon-science","sourceIconUrl":"https://www.amazon.science/apple-touch-icon.png","sourceUrl":"https://www.amazon.science","sourceTier":1},{"id":8856870,"title":"The Gates Foundation is spending $400 million on AI in schools. Teachers warn it could 'end up widening' the divide it's meant to close","titleKo":"Gates Foundation Invests $400 Million in AI Education","hookTitleKo":null,"thumbnailUrl":null,"url":"https://www.reddit.com/r/ArtificialInteligence/comments/1wmjwmj/the_gates_foundation_is_spending_400_million_on/","canonicalUrl":"https://www.reddit.com/r/ArtificialInteligence/comments/1wmjwmj/the_gates_foundation_is_spending_400_million_on","description":"$38","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T17:46:34+00:00","publishedToSiteAt":"2026-09-21T18:08:10.056442+00:00","createdAt":"2026-09-21T18:07:49.257919+00:00","sourceId":57,"isAdminPick":false,"sourceName":"Reddit","sourceSlug":"reddit","sourceIconUrl":null,"sourceUrl":"https://www.reddit.com","sourceTier":2},{"id":8943956,"title":"신뢰로 완성하는 금융 AI의 미래, KodeRunner 2026","titleKo":"KakaoBank Hosts KodeRunner 2026, Presenting Agentic AI Vision Centered on 'Reliability'","hookTitleKo":null,"thumbnailUrl":"https://tech.kakaobank.com/posts/2609-koderunner2026-reliability/images/1-view-of-the-event-venue-lobby.jpg","url":"https://tech.kakaobank.com/posts/2609-koderunner2026-reliability/","canonicalUrl":"https://tech.kakaobank.com/posts/2609-koderunner2026-reliability","description":"2022년 시작된 카카오뱅크의 사내 기술 컨퍼런스 KodeRunner가 2026년 8월, 코엑스 그랜드볼룸에서 다섯 번째 막을 올렸습니다. 올해는 ‘기술적 성장과 콘텐츠’라는 본질에 집중하기 위해 장소를 확장했습니다. ‘Reliability(신뢰성)‘라는 핵심 키 메시지 아래 진행된 키노트, 글로벌 빅테크 4사(MS·Google·AWS·OpenAI) 핸즈온 실습, 마음을 돌보는 Soft Skill 트랙, 그리고 30대의 가챠 머신과 코인교환소, Swap Space로 채워진 로비 이벤트까지, AI 시대를 마주한 카카오뱅크 기술 조직의 치열한 고민과 성장의 기록을 전합니다.","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T17:15:50+00:00","publishedToSiteAt":"2026-09-23T00:05:04.896964+00:00","createdAt":"2026-09-23T00:02:55.986779+00:00","sourceId":105,"isAdminPick":true,"sourceName":"KakaoBank","sourceSlug":"kakaobank-tech","sourceIconUrl":"https://github.com/kakaobank.png","sourceUrl":"https://tech.kakaobank.com/","sourceTier":1},{"id":8945272,"title":"수학과 인공지능 자문그룹","titleKo":"Terence Tao Announces Establishment of AGMAI, a Mathematical Advisory Group for AI Companies","hookTitleKo":null,"thumbnailUrl":"https://terrytao.wordpress.com/wp-content/uploads/2026/09/cropped-thurston-1.png","url":"https://terrytao.wordpress.com/2026/09/21/advisory-group-on-mathematics-and-artificial-intelligence/","canonicalUrl":"https://terrytao.wordpress.com/2026/09/21/advisory-group-on-mathematics-and-artificial-intelligence","description":"

AI 기업과 수학계의 소통, 수학 연구 결과의 책임 있는 발표와 공개를 자문하는 독립 그룹이 프린스턴의 Institute for Advanced Study를 거점으로 출범함
현재는 OpenAI가 내부 모델로 산출했다고 밝힌 다수의 중요한 수학 연구 결과를 어떻게 조율해 공개할지 자...
","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T17:05:51+00:00","publishedToSiteAt":"2026-09-23T00:27:38.860505+00:00","createdAt":"2026-09-23T00:26:58.388581+00:00","sourceId":12,"isAdminPick":false,"sourceName":"GeekNews","sourceSlug":"geeknews","sourceIconUrl":"https://news.hada.io/apple-touch-icon.png","sourceUrl":"https://news.hada.io","sourceTier":2},{"id":8853823,"title":"Show HN: Foremerge – Catch Intent Conflicts Between Parallel Coding Agents","titleKo":"Foremerge v0.5.0 Released: Rust-Based Open-Source Tool Detects Intent Conflicts Between Parallel AI Coding Agents","hookTitleKo":null,"thumbnailUrl":"https://repository-images.githubusercontent.com/1342084139/287b3c71-0a34-4383-a6bc-c7137af4d1a6","url":"https://github.com/naw103/foremerge","canonicalUrl":"https://github.com/naw103/foremerge","description":"Comments","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T16:22:06+00:00","publishedToSiteAt":"2026-09-21T16:54:17.984483+00:00","createdAt":"2026-09-21T16:52:34.083452+00:00","sourceId":11,"isAdminPick":true,"sourceName":"Hacker News","sourceSlug":"hn","sourceIconUrl":"https://news.ycombinator.com/y18.svg","sourceUrl":"https://news.ycombinator.com","sourceTier":2},{"id":8852647,"title":"AWS Weekly Roundup: AWS Builder Center mobile apps, Amazon Connect Talent GA, Amazon Corretto 27, and more (September 21, 2026)","titleKo":"AWS Launches Kimi K3 on Bedrock GA, Releases Corretto 27, and Improves EC2 T8i Performance","hookTitleKo":null,"thumbnailUrl":"https://d2908q01vomqb2.cloudfront.net/da4b9237bacccdf19c0760cab7aec4a8359010b0/2026/09/20/2026-09-21-WIR.png","url":"https://aws.amazon.com/blogs/aws/aws-weekly-roundup-aws-builder-center-mobile-apps-amazon-connect-talent-ga-amazon-corretto-27-and-more-september-14-2026/","canonicalUrl":"https://aws.amazon.com/blogs/aws/aws-weekly-roundup-aws-builder-center-mobile-apps-amazon-connect-talent-ga-amazon-corretto-27-and-more-september-14-2026","description":"Living in the Netherlands, I spend a fair amount of time on trains, and that is usually where I catch up on what the builder community is writing. Until now, that meant opening a laptop or squinting at a browser tab on my phone. This week I found myself scrolling through trending articles and checking […]","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T16:11:24+00:00","publishedToSiteAt":"2026-09-21T16:28:08.168889+00:00","createdAt":"2026-09-21T16:24:38.079198+00:00","sourceId":48,"isAdminPick":false,"sourceName":"AWS Blog","sourceSlug":"aws-blog","sourceIconUrl":null,"sourceUrl":"https://aws.amazon.com/blogs/aws/","sourceTier":1},{"id":8852848,"title":"Why Deploying Physical AI at Scale Demands Safety at Every Layer","titleKo":"NVIDIA Unveils 'Halos', a Full-Stack System for Physical AI Safety","hookTitleKo":null,"thumbnailUrl":"https://blogs.nvidia.com/wp-content/uploads/2026/09/robotics-pr-halos-1600x900-1.png","url":"https://blogs.nvidia.com/blog/physical-ai-halos-safety/","canonicalUrl":"https://blogs.nvidia.com/blog/physical-ai-halos-safety","description":"Physical AI is moving rapidly from research to large-scale deployment. By 2035, ABI Research projects an installed base of 49 million level 3-5 autonomous vehicles (AVs), while Omdia estimates that roughly 60 million industrial robots will be deployed between 2026 and 2035. As these machines enter roads, factories, warehouses and other environments shared with people, […]","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T16:00:46+00:00","publishedToSiteAt":"2026-09-21T16:29:19.813202+00:00","createdAt":"2026-09-21T16:26:46.25753+00:00","sourceId":8,"isAdminPick":false,"sourceName":"NVIDIA Blog","sourceSlug":"nvidia-blog","sourceIconUrl":null,"sourceUrl":"https://blogs.nvidia.com","sourceTier":1},{"id":8852849,"title":"From Enablement to Execution, Egypt’s AI Ecosystem Reaches Production Scale","titleKo":"Egypt and Africa Enter Production Phase with Expanded AI Infrastructure Investment","hookTitleKo":null,"thumbnailUrl":"https://blogs.nvidia.com/wp-content/uploads/2026/09/5696800-ent-egypt-ecosystem-blog-image.png","url":"https://blogs.nvidia.com/blog/egypt-africa-ai-ecosystem/","canonicalUrl":"https://blogs.nvidia.com/blog/egypt-africa-ai-ecosystem","description":"Today, Egypt’s AI builders gathered in the Grand Egyptian Museum for a reception that highlighted the nation’s rapidly growing AI ecosystem — spanning AI natives, developers, researchers, startups and enterprises — building applications across industries. The event included a keynote from Paolo Guglielmini, vice president of EMEA at NVIDIA. Ahmed Mostafa, regional AI adoption lead […]","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T16:00:10+00:00","publishedToSiteAt":"2026-09-21T16:29:11.704423+00:00","createdAt":"2026-09-21T16:26:46.25753+00:00","sourceId":8,"isAdminPick":false,"sourceName":"NVIDIA Blog","sourceSlug":"nvidia-blog","sourceIconUrl":null,"sourceUrl":"https://blogs.nvidia.com","sourceTier":1},{"id":9174684,"title":"Loopjacking in A2A Implementations: Hijacking Human-in-the-Loop Approvals","titleKo":"LangGraph Agent Server Vulnerability Allows Approval Bypass via State Mutation","hookTitleKo":null,"thumbnailUrl":"https://adithyanak.com/images/posts/loopjacking-a2a-task-update-cover.png","url":"https://adithyanak.com/loopjacking-in-a2a-implementations/","canonicalUrl":"https://adithyanak.com/loopjacking-in-a2a-implementations","description":"Comments","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T16:00:00+00:00","publishedToSiteAt":"2026-09-25T23:45:00.631042+00:00","createdAt":"2026-09-25T23:43:13.529216+00:00","sourceId":11,"isAdminPick":false,"sourceName":"Hacker News","sourceSlug":"hn","sourceIconUrl":"https://news.ycombinator.com/y18.svg","sourceUrl":"https://news.ycombinator.com","sourceTier":2},{"id":8920681,"title":"AWS Launches Strands Harness, an Agent That Brings Its Own Everything but the Model (6 minute read)","titleKo":"AWS Releases 'Strands Harness' Based on Strands Agents SDK with Complete Agent Features Excluding Model Intelligence","hookTitleKo":null,"thumbnailUrl":"https://thelettertwo.com/wp-content/uploads/2024/12/aws-logo-reinvent-2024-kenyeung-960x540-1.jpg","url":"https://thelettertwo.com/2026/09/21/aws-strands-harness?utm_source=tldrai","canonicalUrl":"https://thelettertwo.com/2026/09/21/aws-strands-harness","description":"AWS' new Strands harness is a ready-to-run agent that can search the web, run commands, edit files, remember what it did yesterday, and hand off work to helper agents.","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T16:00:00+00:00","publishedToSiteAt":"2026-09-22T17:05:23.388619+00:00","createdAt":"2026-09-22T16:55:56.65085+00:00","sourceId":14,"isAdminPick":false,"sourceName":"TLDR AI","sourceSlug":"tldr-ai","sourceIconUrl":"https://tldr.tech/apple-touch-icon.png","sourceUrl":"https://tldr.tech/ai","sourceTier":2},{"id":8854806,"title":"Jev is now available in LangSmith Evals","titleKo":"LangSmith Adds Jev Evaluation Model… Agent Evaluation Up to 450x Cheaper and 200x Faster Than LLMs","hookTitleKo":null,"thumbnailUrl":"https://cdn.prod.website-files.com/65c81e88c254bb0f97633a71/6ab0bd7112e026354354705b_header-green-dark.png","url":"https://www.langchain.com/blog/jev-is-now-available-in-langsmith-evals","canonicalUrl":"https://www.langchain.com/blog/jev-is-now-available-in-langsmith-evals","description":"Use Jev as a judge for LangSmith evals to evaluate agent traces with faster, cheaper structured feedback across production runs, datasets, and regression tests.","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T16:00:00+00:00","publishedToSiteAt":"2026-09-21T17:19:22.619142+00:00","createdAt":"2026-09-21T17:17:31.139095+00:00","sourceId":10,"isAdminPick":false,"sourceName":"LangChain Blog","sourceSlug":"langchain-blog","sourceIconUrl":null,"sourceUrl":"https://www.langchain.com/blog","sourceTier":1},{"id":8850867,"title":"Improving synthesis prediction of small molecules at scale with RetroChimera","titleKo":"Nature-published retrosynthesis AI 'RetroChimera' released… Demonstrates superior performance over existing models in expert evaluation","hookTitleKo":null,"thumbnailUrl":"https://www.microsoft.com/en-us/research/wp-content/uploads/2026/09/RetroChimera-BlogHeroFeature-1400x788-1-1280x720.jpg","url":"https://www.microsoft.com/en-us/research/blog/improving-synthesis-prediction-of-small-molecules-at-scale-with-retrochimera/","canonicalUrl":"https://www.microsoft.com/en-us/research/blog/improving-synthesis-prediction-of-small-molecules-at-scale-with-retrochimera","description":"
Custom-made molecules are advancing medicine, materials, and agriculture, but producing them is slow and expensive. A new Nature paper highlights RetroChimera, a predictive model that helps accelerate chemical synthesis, helping researchers explore a wide range of molecules.
\n
The post Improving synthesis prediction of small molecules at scale with RetroChimera appeared first on Microsoft Research.
","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T15:30:19+00:00","publishedToSiteAt":"2026-09-21T15:40:03.799494+00:00","createdAt":"2026-09-21T15:38:30.247815+00:00","sourceId":5,"isAdminPick":false,"sourceName":"Microsoft Research","sourceSlug":"ms-research","sourceIconUrl":null,"sourceUrl":"https://www.microsoft.com/en-us/research/","sourceTier":1},{"id":8849150,"title":"GitHub Copilot for JetBrains - v1.18 Updates (Org-level skills, Assisted Approvals, Side-by-side view)","titleKo":"GitHub Releases Copilot for JetBrains v1.18 Update","hookTitleKo":null,"thumbnailUrl":null,"url":"https://www.reddit.com/r/GithubCopilot/comments/1wmfcx7/github_copilot_for_jetbrains_v118_updates/","canonicalUrl":"https://www.reddit.com/r/GithubCopilot/comments/1wmfcx7/github_copilot_for_jetbrains_v118_updates","description":"$39","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T15:04:36+00:00","publishedToSiteAt":"2026-09-21T15:07:17.365844+00:00","createdAt":"2026-09-21T15:06:48.858277+00:00","sourceId":57,"isAdminPick":false,"sourceName":"Reddit","sourceSlug":"reddit","sourceIconUrl":null,"sourceUrl":"https://www.reddit.com","sourceTier":2},{"id":8929727,"title":"A study of sequence weighting at scale","titleKo":"Jane Street Research Discovers Non-Monotonic Patterns in Sequence Weight Scaling","hookTitleKo":null,"thumbnailUrl":"https://blog.janestreet.com/a-study-of-sequence-weighting-at-scale/featimg.jpg","url":"https://blog.janestreet.com/a-study-of-sequence-weighting-at-scale/","canonicalUrl":"https://blog.janestreet.com/a-study-of-sequence-weighting-at-scale","description":"Comments","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T15:03:33+00:00","publishedToSiteAt":"2026-09-22T19:41:15.140373+00:00","createdAt":"2026-09-22T19:39:08.251876+00:00","sourceId":11,"isAdminPick":false,"sourceName":"Hacker News","sourceSlug":"hn","sourceIconUrl":"https://news.ycombinator.com/y18.svg","sourceUrl":"https://news.ycombinator.com","sourceTier":2},{"id":8849152,"title":"WarpState: experimenting with dual-timescale associative memory + local attention for long-running LLMs","titleKo":"WarpState Reveals Dual-Memory Architecture for Long-Context LLMs","hookTitleKo":null,"thumbnailUrl":null,"url":"https://www.reddit.com/r/LocalLLaMA/comments/1wmf590/warpstate_experimenting_with_dualtimescale/","canonicalUrl":"https://www.reddit.com/r/LocalLLaMA/comments/1wmf590/warpstate_experimenting_with_dualtimescale","description":"$3a","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T14:57:06+00:00","publishedToSiteAt":"2026-09-21T15:07:46.970201+00:00","createdAt":"2026-09-21T15:06:48.858277+00:00","sourceId":57,"isAdminPick":false,"sourceName":"Reddit","sourceSlug":"reddit","sourceIconUrl":null,"sourceUrl":"https://www.reddit.com","sourceTier":2},{"id":8849959,"title":"AI Security Is an Engineering Problem — How to Solve It at Every Layer of the Agent Stack","titleKo":"AI Security Must Be Treated as an Engineering Problem: Emphasizing Control and Verification Across All Layers of the Agent Stack","hookTitleKo":null,"thumbnailUrl":"https://blogs.nvidia.com/wp-content/uploads/2026/09/security-thought-leadership-1920x1080-1.jpg","url":"https://blogs.nvidia.com/blog/ai-security-agent-stack/","canonicalUrl":"https://blogs.nvidia.com/blog/ai-security-agent-stack","description":"AI security is an engineering problem. That means defined security requirements, enforceable controls, named owners and evidence that protections work. As AI becomes more capable, the industry must accelerate security engineering, broaden access to defensive tools and share what works faster. Technology Changes, Security Fundamentals Endure The internet and cloud computing changed how software operates, […]","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T14:51:34+00:00","publishedToSiteAt":"2026-09-21T15:16:00.124105+00:00","createdAt":"2026-09-21T15:14:43.863479+00:00","sourceId":8,"isAdminPick":false,"sourceName":"NVIDIA Blog","sourceSlug":"nvidia-blog","sourceIconUrl":null,"sourceUrl":"https://blogs.nvidia.com","sourceTier":1},{"id":8847795,"title":"Introducing Our New Dropbox API Documentation","titleKo":"Dropbox Releases New API Documentation with MCP Support and AI Assistant","hookTitleKo":null,"thumbnailUrl":"https://dropbox.tech/cms/content/dam/dropbox/tech-blog/en-us/developers/Developers-1-1440x305px-light.png","url":"https://dropbox.tech/developers/new-dropbox-api-documentation","canonicalUrl":"https://dropbox.tech/developers/new-dropbox-api-documentation","description":"Learn about the newly launched version of the Dropbox API documentation and its modern design and features.","score":0,"commentCount":0,"viewCount":1,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T14:24:44.522832+00:00","publishedToSiteAt":"2026-09-21T14:27:45.109306+00:00","createdAt":"2026-09-21T14:24:44.524168+00:00","sourceId":52,"isAdminPick":true,"sourceName":"Dropbox Tech","sourceSlug":"dropbox-tech","sourceIconUrl":null,"sourceUrl":"https://dropbox.tech","sourceTier":1},{"id":8846890,"title":"Pruning LLMs Like a Physicist: Block Removal as an Ising Optimization Problem","titleKo":"Multiverse Publishes Paper on LLM Block Removal Based on Ising Optimization","hookTitleKo":null,"thumbnailUrl":"https://cdn-uploads.huggingface.co/production/uploads/693c2a4eb0871ba57155b4ed/q4gZV9ENtfFwge7H3Q7YJ.png","url":"https://huggingface.co/blog/MultiverseComputingCAI/pruning-llms-like-a-physicist-block-removal-as-an","canonicalUrl":"https://huggingface.co/blog/MultiverseComputingCAI/pruning-llms-like-a-physicist-block-removal-as-an","description":null,"score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T13:44:34+00:00","publishedToSiteAt":"2026-09-21T14:02:59.871659+00:00","createdAt":"2026-09-21T14:00:22.301711+00:00","sourceId":3,"isAdminPick":false,"sourceName":"HuggingFace Blog","sourceSlug":"huggingface","sourceIconUrl":null,"sourceUrl":"https://huggingface.co/blog","sourceTier":1},{"id":8845594,"title":"Key-independent Collisions in Hash Functions at Scale using Claude and Codex","titleKo":"AI-Based Hash Collision Analysis Confirms No Performance-Security Tradeoff","hookTitleKo":null,"thumbnailUrl":null,"url":"https://www.reddit.com/r/OpenAI/comments/1wmcnx1/keyindependent_collisions_in_hash_functions_at/","canonicalUrl":"https://www.reddit.com/r/OpenAI/comments/1wmcnx1/keyindependent_collisions_in_hash_functions_at","description":"AI tools can now be used to mass analyze hash functions for collisions, but also for proving correctness in Lean. We discovered collisions in komihash, a5hash, HighwayHash, SpookyHash, aHash, and t1ha2 including the standard hashes in Rust, Go, and .NET. We then plotted the collision resistance against the hash speed. And it turns out there's not actually a tradeoff.","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T13:22:01+00:00","publishedToSiteAt":"2026-09-21T13:37:50.802906+00:00","createdAt":"2026-09-21T13:37:33.574072+00:00","sourceId":57,"isAdminPick":false,"sourceName":"Reddit","sourceSlug":"reddit","sourceIconUrl":null,"sourceUrl":"https://www.reddit.com","sourceTier":2},{"id":8844343,"title":"Python Workers are now generally available","titleKo":"Cloudflare Officially Launches Python Workers... Supporting FastAPI, Django, and AI Pipeline Construction","hookTitleKo":null,"thumbnailUrl":"https://blog.cloudflare.com/_emdash/api/media/file/01M2V34BA05P39V3QCVSXF412X.01M2V34C4FZEN00M4XFVXH4W2K.png","url":"https://blog.cloudflare.com/python-workers-ga/","canonicalUrl":"https://blog.cloudflare.com/python-workers-ga","description":"Python Workers allow developers to run Python web frameworks and AI orchestration libraries natively in the Cloudflare Workers runtime. You can seamlessly integrate with Cloudflare's ecosystem including D1, R2, and Workers AI without writing any JavaScript glue code.","score":0,"commentCount":0,"viewCount":0,"bookmarkCount":0,"likeCount":0,"dislikeCount":0,"externalSource":null,"state":"processed","aiCategory":null,"summaryStatus":"done","publishedAt":"2026-09-21T13:00:00+00:00","publishedToSiteAt":"2026-09-21T13:12:15.146894+00:00","createdAt":"2026-09-21T13:08:39.528812+00:00","sourceId":46,"isAdminPick":false,"sourceName":"Cloudflare Blog","sourceSlug":"cloudflare-blog","sourceIconUrl":null,"sourceUrl":"https://blog.cloudflare.com","sourceTier":1}],"total":14843}}]

Latest AI news

Amazon Launches AI, Energy, and Healthcare Research Initiative with Stanford

OpenAI and Anthropic's RSI Strategy and the Uncertainty of AI Profitability

Meta's 'Muse' Surpasses 2.5 Million Downloads Two Weeks After Launch, Overtaking ChatGPT on iOS

NVIDIA Unveils 'DSX Ready' Certification Program for Power and Cooling Products for AI Factories

Amazon Releases Three Papers on AI-Based Antibody Design and Evaluation

Gates Foundation Invests $400 Million in AI Education

PickKakaoBank Hosts KodeRunner 2026, Presenting Agentic AI Vision Centered on 'Reliability'

Terence Tao Announces Establishment of AGMAI, a Mathematical Advisory Group for AI Companies

PickForemerge v0.5.0 Released: Rust-Based Open-Source Tool Detects Intent Conflicts Between Parallel AI Coding Agents

AWS Launches Kimi K3 on Bedrock GA, Releases Corretto 27, and Improves EC2 T8i Performance

NVIDIA Unveils 'Halos', a Full-Stack System for Physical AI Safety

Egypt and Africa Enter Production Phase with Expanded AI Infrastructure Investment

NEWLangGraph Agent Server Vulnerability Allows Approval Bypass via State Mutation

AWS Releases 'Strands Harness' Based on Strands Agents SDK with Complete Agent Features Excluding Model Intelligence

LangSmith Adds Jev Evaluation Model… Agent Evaluation Up to 450x Cheaper and 200x Faster Than LLMs

Nature-published retrosynthesis AI 'RetroChimera' released… Demonstrates superior performance over existing models in expert evaluation

GitHub Releases Copilot for JetBrains v1.18 Update

Jane Street Research Discovers Non-Monotonic Patterns in Sequence Weight Scaling

WarpState Reveals Dual-Memory Architecture for Long-Context LLMs

AI Security Must Be Treated as an Engineering Problem: Emphasizing Control and Verification Across All Layers of the Agent Stack

PickDropbox Releases New API Documentation with MCP Support and AI Assistant

Multiverse Publishes Paper on LLM Block Removal Based on Ising Optimization

AI-Based Hash Collision Analysis Confirms No Performance-Security Tradeoff

Cloudflare Officially Launches Python Workers... Supporting FastAPI, Django, and AI Pipeline Construction