AI News: Community https://ai-news.ypetit.net/sources/ Community publications from the world's largest AI labs and companies, aggregated by AI News. Primary sources only. en Fri, 23 Oct 2026 19:08:30 GMT Latent Space (swyx): AI Is Getting Good at Math. Here’s Why.https://latent.space/p/ai-is-getting-good-at-math/urn:latent-space:post:post_46e4845e-1c40-414a-8538-c0bf636fb11aFri, 23 Oct 2026 15:19:48 GMTcommunityAI models are solving more mathematical problems. A pattern is emerging in their successes that could help us predict which problems they’ll solve next. The Batch (Andrew Ng): A Single Decoder Turns Small Models into Action: TurboVLA turns to a throwback architecture to attend to language and visionhttps://www.deeplearning.ai/the-batch/a-single-decoder-turns-small-models-into-action/6ad930af04046100012d200cFri, 23 Oct 2026 15:15:00 GMTcommunityThe vision-language-action architecture that increasingly drives robots relies on a large language model to combine visual observations with text instructions before an action decoder produces motor commands. The Batch (Andrew Ng): Nvidia Courts A New Partner: Reflection releases Beam, an open weight model from an independent U.S.-based labhttps://www.deeplearning.ai/the-batch/nvidia-courts-a-new-partner/6ad9300904046100012d2005Fri, 23 Oct 2026 15:10:00 GMTcommunityReflection AI, a closely-watched, Nvidia-backed U.S. startup, announced its first model. The Batch (Andrew Ng): ChatGPT for Teens Gets a Report Card: OpenAI adds features for youths as studies doubt program's safety and educational impacthttps://www.deeplearning.ai/the-batch/chatgpt-for-teens-gets-a-report-card/6ad92a659f1f3400019b1b50Fri, 23 Oct 2026 15:05:00 GMTcommunityOpenAI announced that it would add college-planning and study tools to its under-18 experience. The Batch (Andrew Ng): Meta Releases Wonder, Opens Spark: All about Muse Wonder (API), Muse Wonder Full (subscription), and Muse Spark Open (weights available)https://www.deeplearning.ai/the-batch/meta-releases-wonder-opens-spark/6ad926ff9f1f3400019b1b42Fri, 23 Oct 2026 15:00:00 GMTcommunityMeta marched all the way back to the AI frontier to offer a Muse model for everyone. The Batch (Andrew Ng): The Four Ps — A Leadership Framework: How to Manage An AI-Native Engineering Teamhttps://www.deeplearning.ai/the-batch/the-four-ps/6ada6f539d8d710001dec52aFri, 23 Oct 2026 14:30:00 GMTcommunityWhat should an engineering leader such as a VP Engineering or an Engineering manager do to help their team become more proficient in AI? The Batch (Andrew Ng): Meta’s Muse Wonder and Spark Open, Educational Use of ChatGPT, Reflection's Partnership with Nvidiahttps://www.deeplearning.ai/the-batch/issue-376/6ada6f35a438b50001a256b0Fri, 23 Oct 2026 14:00:00 GMTcommunityThe Batch News & Insights: What should an engineering leader such as a VP Engineering or an Engineering manager do to help their team become more proficient in AI? Latent Space (swyx): [AINews] Black Octoberhttps://latent.space/p/ainews-black-october/urn:latent-space:post:post_3049d28e-7e21-4f4e-9f4b-ff63b89ebd0dFri, 23 Oct 2026 10:09:29 GMTcommunitythe Cyber onslaught continues. Latent Space (swyx): Why AI Compute Comes in Rectangles — Erik Bernhardsson, Modal CEOhttps://latent.space/p/modal-d/https://www.latent.space/p/modal-dThu, 22 Oct 2026 17:40:57 GMTcommunityElastic GPUs, sandbox growth, and the economics of the agent cloud Latent Space (swyx): You are the Team — Anekha Sokhal, Moshihttps://latent.space/p/moshi/https://www.latent.space/p/moshiThu, 22 Oct 2026 14:03:40 GMTcommunityHow Moshi turns market data into actionable insights, and why AI tools have to fit the trader’s workflow. Latent Space (swyx): [AINews] OpenAI’s Patch the Planet program is 1 of 2 parties to discover PHP2Shell vulnerabilities in core PHP affecting 70% of the web, using GPT-5.6 Sol; coordinated response over 2 monthshttps://latent.space/p/ainews-openais-patch-the-planet-program/https://www.latent.space/p/ainews-openais-patch-the-planet-programThu, 22 Oct 2026 07:42:44 GMTcommunityThe Big Hardening begins The Batch (Andrew Ng): Minimax’s M3.1 Flash Cuts Prices: Critical PHP bug found by AI affects majority of web siteshttps://www.deeplearning.ai/the-batch/minimaxs-m3-1-flash-cuts-prices/6ada49c89d8d710001dec511Wed, 21 Oct 2026 20:40:00 GMTcommunityMercury 3, diffusion-based alternative to Claude Haiku. d1-ultra, a highly accurate decision model. Using OpenAI’s models to rewrite and repair legacy code. llm-d, an all-around optimizer for serving open models. Latent Space (swyx): Claude Mods: Flipping the Harness Inside Outhttps://latent.space/p/claude-mods/https://www.latent.space/p/claude-modsWed, 21 Oct 2026 13:04:19 GMTcommunityClaude Mods is an early form of the Attention Interface. It flips the agent harness paradigm from the harness facing the machine, to the harness facing the human. Latent Space (swyx): [AINews] DeepSeek DeepInfer: Open Source SOTA Inference running 155T toks/day/cluster making $22m/day at ~80% margin for a 81 day paybackhttps://latent.space/p/ainews-deepseek-deepinfer-open-source/https://www.latent.space/p/ainews-deepseek-deepinfer-open-sourceWed, 21 Oct 2026 06:54:09 GMTcommunityAn open source “bomb” drops from China Latent Space (swyx): AI for the Fortune 2000: How Tenex Builds an AI-Native SDLC — Alex Lieberman & CJ Hesshttps://latent.space/p/tenex/https://www.latent.space/p/tenexWed, 21 Oct 2026 05:01:26 GMTcommunityPaying for outcomes, building with agents, and the last mile of enterprise AI adoption Latent Space (swyx): The Programmable Data Center — Evan Conrad, SF Computehttps://latent.space/p/sfcompute2/https://www.latent.space/p/sfcompute2Tue, 20 Oct 2026 19:31:57 GMTcommunityFrom GPU trading to physical settlement, independent markets, and programmable compute Latent Space (swyx): [AINews] not much happened todayhttps://latent.space/p/ainews-not-much-happened-today-f09/https://www.latent.space/p/ainews-not-much-happened-today-f09Tue, 20 Oct 2026 05:21:52 GMTcommunityCongrats to DeepSeek V4.1 Flash 1017, MiniMax M3.1 Flash, and Mercury 3 (details below) but they’re not going to be in super wide use just yet. The Batch (Andrew Ng): Muse Spark Open leads U.S. open models: Reflection releases weights for Beam, a much-anticipated modelhttps://www.deeplearning.ai/the-batch/muse-spark-open-leads-u-s-open-models/6ada486c9d8d710001dec4ffMon, 19 Oct 2026 21:35:00 GMTcommunityGrok Voice 3.0 leads voice-to-voice benchmarks. Nvidia 3.5 Super VL is faster, more capable than Lightning. DeepSeek-V-4.1-Flash pushes strong update. Meta amends controversial Muse scraping terms. Latent Space (swyx): [AINews] High Taste Testers Ep. 1https://latent.space/p/ainews-high-taste-testers-ep-1/https://www.latent.space/p/ainews-high-taste-testers-ep-1Sat, 17 Oct 2026 07:47:38 GMTcommunitya quiet day lets us pilot a new show The Batch (Andrew Ng): Data Mixing, Minus the Guesswork: Research from the team behind CausalMix offers a technique to select fine-tuning exampleshttps://www.deeplearning.ai/the-batch/data-mixing-minus-the-guesswork/6ace5ef23a889d0001158d8cFri, 16 Oct 2026 15:15:44 GMTcommunityWhen you’re building a fine-tuning dataset, finding a mix of subject domains that yields a high-performing model can require expensive trial and error, because fine-tuned performance also depends on the quality of the individual examples. The Batch (Andrew Ng): Computers Built for Local AI: Apple’s Mac mini and Mac Studio have competition from Microsoft and Nvidiahttps://www.deeplearning.ai/the-batch/computers-built-for-local-ai/6ace5e0be1b17c0001fa49f0Fri, 16 Oct 2026 15:10:27 GMTcommunityDevelopers have warmed to the economics and security of local models, either as an alternative or a complement to the cloud. The Batch (Andrew Ng): Mistral’s Sovereign Appeal: Mistral Large 4, the leading European AI model, outperforms all open models outside of Chinahttps://www.deeplearning.ai/the-batch/mistrals-skills-match-its-global-appeal/6ace5c4fe1b17c0001fa49e0Fri, 16 Oct 2026 15:05:12 GMTcommunityMistral’s latest model is its largest and most capable, and the most powerful yet released by a European company. The Batch (Andrew Ng): Claude Fills Out Its Model Family: Claude Sonnet 5.5 and Haiku 5.5 give developers fast, highly capable alternatives to Opus 5.5https://www.deeplearning.ai/the-batch/claude-fills-out-its-model-family/6ace5f2b3a889d0001158d90Fri, 16 Oct 2026 15:00:58 GMTcommunityClaude Haiku, Anthropic’s fastest and least expensive model gets a long-awaited update. The Batch (Andrew Ng): Formal Verification Meets Unverifiable Judgement: Don’t Overgeneralize from Progress in Math and Code — The Hardest AI Problems Still Require More Workhttps://www.deeplearning.ai/the-batch/formal-verification-meets-unverifiable-judgement/6ad13476eb5c2b000165d5c2Fri, 16 Oct 2026 14:30:41 GMTcommunityThe gulf between LLM performance on verifiable tasks vs. unverifiable tasks that require judgement is larger than most people realize. The Batch (Andrew Ng): Claude Haiku 5.5, Mistral Large 4, and Local Models for Mac and Windowshttps://www.deeplearning.ai/the-batch/issue-375/6ad136eceb5c2b000165d5ddFri, 16 Oct 2026 14:00:12 GMTcommunityThe Batch News & Insights: The gulf between LLM performance on verifiable tasks vs. unverifiable tasks that require judgement is larger than most people realize. Latent Space (swyx): [AINews] Agents disprove Polysemanticity; greatly improve Interpretability & Alignmenthttps://latent.space/p/ainews-agents-disprove-polysemanticity/https://www.latent.space/p/ainews-agents-disprove-polysemanticityFri, 16 Oct 2026 04:16:45 GMTcommunityNick Cammarata out of nowhere with some major claims! Latent Space (swyx): [AINews] Modal’s $1B ARR, $16B Series D makes it NYC’s newest AI Infra Decacorn, 3x usage and 3x valuation in 5 months, 15x in 1 yrhttps://latent.space/p/ainews-modals-1b-arr-16b-series-d/https://www.latent.space/p/ainews-modals-1b-arr-16b-series-dThu, 15 Oct 2026 10:08:08 GMTcommunitySandboxes are all you need. Latent Space (swyx): Crossing the Threshold: How OpenAI and Anthropic Approach Autonomy and Safety in Financial Agentshttps://latent.space/p/crossing-the-threshold/https://www.latent.space/p/crossing-the-thresholdWed, 14 Oct 2026 21:40:52 GMTcommunityAt AI Engineer New York, frontier labs and financial firms explored how to give agents useful autonomy while keeping consequential decisions under control. Latent Space (swyx): Lab-in-the-Loop at Compute Speed — Ian Quigley, Leash Biohttps://latent.space/p/lab-in-the-loop-at-compute-speed/https://www.latent.space/p/lab-in-the-loop-at-compute-speedWed, 14 Oct 2026 20:35:09 GMTcommunityHow Leash Bio tamed DELs, the assay that traumatizes everyone, and used the data to catch cheating models. And why that same data is perfect for AI agents learning to think like chemists. The Batch (Andrew Ng): Meta’s Muse Wonder rivals top models: Inside Amazon’s warehouse robotics operationshttps://www.deeplearning.ai/the-batch/metas-muse-wonder-rivals-top-models/6ad0fbb882061100014c6347Wed, 14 Oct 2026 19:17:00 GMTcommunityMicrosoft says it will use own chips for Copilot, MAI. Gemini 4 Argon generally available. Python 3.15 is official, Python 3.10 no longer supported. Character.AI declares bankruptcy. Latent Space (swyx): [AINews] AI Engineer NYC: Gemini 4 Argon, Vals AI Benchmarkshttps://latent.space/p/ainews-ai-engineer-nyc-gemini-4-argon/https://www.latent.space/p/ainews-ai-engineer-nyc-gemini-4-argonWed, 14 Oct 2026 05:40:28 GMTcommunityGemini 4 Argon went GA today, but we already gave it title story 2 weeks ago and most benchmarks, like CursorBench, simply independently confirm their claims. Latent Space (swyx): Dyna-2: General Purpose Physical Agents with Hour-Long Autonomy — Jason Ma, Dyna Roboticshttps://latent.space/p/dyna/https://www.latent.space/p/dynaTue, 13 Oct 2026 22:25:55 GMTcommunityOur first robotics pod with Dyna’s Chief Scientist on their big bet on World Action Models & the Next Frontier of Physical AI Latent Space (swyx): [AINews] Meta Muse Wonder & Wonder Full: MSL’s Claude 5.5-tier Frontier Modelhttps://latent.space/p/ainews-meta-muse-wonder-and-wonder/https://www.latent.space/p/ainews-meta-muse-wonder-and-wonderTue, 13 Oct 2026 09:37:04 GMTcommunityMSL’s emergence as a frontier lab is finally complete. Latent Space (swyx): How Anthropic’s Dynamic Workflows Coordinate Claude Agents (Updated with video!)https://latent.space/p/claude-dynamic-workflows/https://www.latent.space/p/claude-dynamic-workflowsMon, 12 Oct 2026 18:18:17 GMTcommunityAt AI Engineer New York, Anthropic drew on lessons from Unix and Linux to explain how agents can tackle bigger projects without overwhelming developers. Latent Space (swyx): [Subscriber Exclusive] NYC Subscriber Meetups!https://latent.space/p/nyc2026/https://www.latent.space/p/nyc2026Sat, 10 Oct 2026 22:07:17 GMTcommunityTomorrow and Tuesday. If you see this you’re in!