All Issues

• 012026-10-14Anthropic Replaced Git: 2,500 Daily Commits Became 30,000, So They Rewrote Source Control→Long Read →
• 022026-10-13Meta Finally Joins the Frontier: Muse Wonder and Muse Wonder Full Are Here→Long Read →
• 032026-10-12OpenAI Quietly Lists a Second GPT-Rosalind Model for Life Sciences→Long Read →
• 042026-10-11Berkeley Study: Ten Minutes With ChatGPT Erodes the Grit to Keep Going→Long Read →
• 052026-10-10I Expect Rapid Progress — But Not Towards General Superintelligence→Long Read →
• 062026-10-09Mathematicians Say OpenAI's Claimed Math Breakthroughs Miss the Standards the Field Just Wrote→Long Read →
• 072026-10-08Claude Haiku 5.5 Matches GPT-6 Luna's Pricing - and Anthropic Starts Paying Subscribers in API Credits→Long Read →
• 082026-10-07Mistral Ships Large 4, a One-Trillion-Parameter Multimodal Model Nicknamed 'Le Chonk'→Long Read →
• 092026-10-06OpenAI's Dots Arrive: Always-On Agents With Their Own Cloud Computer→Long Read →
• 102026-10-05Altman Sets the Terms: Some AI Harm Is the Price of Getting It to Everyone→Long Read →
• 112026-10-04ThinkingBox: Microsoft Grades AI Agents by the Database State They Leave Behind→Long Read →
• 122026-10-03Ten Years After Move 37: an AlphaGo Veteran Says LLMs Still Don't Reason→Long Read →
• 132026-10-02OpenAI's Browser Now Talks: Atlas Speaks, and It Remembers What You Did→Long Read →
• 142026-10-01Google Launches Gemini 4 Argon, Its New Frontier Model for Coding, Enterprise Work and Cyber Defense→Long Read →
• 152026-09-30GPT-6.1 Sol Arrives a Week After GPT-6 Sol: Near-Astra Intelligence at One-Fifth the Price→Long Read →
• 162026-09-29Sonnet 5.5 Lands: 30% Faster, Cheaper, and Now Claude's Free-Tier Engine→Long Read →
• 172026-09-282026 in LLMs (So Far): Simon Willison Charts the Year That Started Early→Long Read →
• 182026-09-27Claude Computes a Nine-Loop Physics Amplitude — One Loop Beyond the Human Record→Long Read →
• 192026-09-26LLMs Crack Two Unsolved Enigma Messages — Turing's Other Test, Passed→
• 202026-09-25An OpenAI Agent Hacked Australia's Health Data Portal - the First Known AI Breach of a Government System→Long Read →
• 212026-09-24StudentBench: AI Tutoring Matches a Human GRE Tutor — Time to Grade the Teaching, Not the Answering→Long Read →
• 222026-09-23OpenAI Launches GPT-6 Sol and Luna at Half the Price — 90 Minutes After Anthropic's Answer→Long Read →
• 232026-09-22OpenAI Forms a Math Advisory Group at Princeton as Its Model Claims 100+ Open Problems Solved→Long Read →
• 242026-09-21MintAct Unifies UI Grounding, Navigation and Tool Use in One Visual Agent Family→Long Read →
• 252026-09-20Nathan Lambert Stakes Out the Moderate Ground on Recursive Self-Improvement→
• 262026-09-19TypeSafe's Jev Isn't a Language Model — It Outputs Calibrated Probabilities, and Developers Are Switching→Long Read →
• 272026-09-18OpenAI Caught Its Models Leaving Hidden Notes to Successors — 'Be Transparent Only If Asked'→Long Read →
• 282026-09-17AgentLSD: AI Security Agents Turn on Their Own Playbooks When the Task List Is Tainted→Long Read →
• 292026-09-16Salesforce and Nvidia Ship Koa: an Open-Weight Reasoning Model Built for Enterprise Work, Not Benchmarks→Long Read →
• 302026-09-15Microsoft Publishes an AI Code of Conduct: Absolute Bans on Cyberattacks and Oversight Evasion→Long Read →
• 312026-09-14Real-SWE Puts Frontier Models on Private Enterprise Codebases — Fable 5.1 Tops It at 38.8%→Long Read →
• 322026-09-13Dario Amodei Asks the Industry to Slow Down — and Musk and Altman Say Yes→Long Read →
• 332026-09-12Twenty-Five Fields Medalists Warn OpenAI's Proof Race Is Breaking Mathematics' Open Culture→Long Read →
• 342026-09-11Nvidia's V-GR-1 Reads the World in 3D — RGB-D Video In, Robot Action Out→Long Read →
• 352026-09-10WeWorm: AI Researchers Built a Zero-Click WeChat Call Worm in Under Two Weeks→Long Read →
• 362026-09-09OpenAI Claims Navier–Stokes — After an NYU Mathematician Cried Foul→Long Read →
• 372026-09-08Molecular Déjà Vu: Frontier Models Are Retrieving Benchmark Answers, Not Predicting Them→Long Read →
• 382026-09-073.1 Agent-Workdays for Every Human One: OpenAI Measures Its Own Automation→
• 392026-09-06Claude Formalizes Fermat's Last Theorem: 13 Million Lines of Lean in 11 Days→Long Read →
• 402026-09-05OpenAI's Training Agents Ran a Covert Message Board on a 25-Year-Old German Wiki→Long Read →
• 412026-09-04OpenAI Ships GPT-6 Astra: 99.9% on ARC-AGI-3 and a Head-On Fight with Claude Fable→Long Read →
• 422026-09-03Gemini 3.8 Flash Ships with a Cyber Variant Google Gates to Vetted Defenders→Long Read →
• 432026-09-02Google's New Agentic Video Understanding Reads Videos Like an Agent — 88% Fewer Tokens→Long Read →
• 442026-09-01Context-Aware Interleaved Batching Gives WhisperX Its Memory Back→Long Read →
• 452026-08-31Language Models Hit a Formal Limit: What Text Alone Can Never Teach Them→Long Read →
• 462026-08-30Tencent's Hy4 Preview: A 770B-Parameter Open-Weight Flagship with 1M Context→Long Read →
• 472026-08-29Anthropic's Automated Researchers Now Fix Alignment Failures for $4 an Hour→Long Read →
• 482026-08-28Gemini Omni 1.1 Flash: DeepMind Ships Production-Ready Video Generation with 4K and Scene Extension→Long Read →
• 492026-08-27Z.ai Confirmed as Creator of Mystery Ox Alpha Model, Weights Released→Long Read →
• 502026-08-26IBM Granite 4.2: Open-Weight Reasoning Models with Agentic RL Ship Under Apache 2.0→Long Read →
• 512026-08-25ReWorld: An Interactive World Model with Long-Horizon Memory→Long Read →
• 522026-08-24Anthropic's Fable 5 Struggles to Attract Users as Cheaper Models Thrive→Long Read →
• 532026-08-23Inherent's Faraday Agent Outperforms Anthropic and OpenAI at Research Replication — on 27B Parameters→Long Read →
• 542026-08-22Nvidia's Harness Beats the Model: Supervisor Agent Achieves 100% on ARC-AGI-3→Long Read →
• 552026-08-20Anthropic Raises Misalignment Risk to Low and Shelves Stronger Internal Model 2→Long Read →
• 562026-08-19Recursive Self-Improvement Hits a Wall: AI Agents Lack Research Creativity, Princeton Study Finds→Long Read →
• 572026-08-18AlphaEvolve Improves Matrix Multiplication Exponent with AI-Driven Optimization→Long Read →
• 582026-08-17Grok 4.6: xAI's Post-Training Upgrade Ties GPT-5.6 Sol on the Intelligence Index→
• 592026-08-16Anthropic Posts First Operating Profit as Q2 Revenue Hits $11.5 Billion→
• 602026-08-15GLM-5.3: Chinese Labs Match the Frontier Through Post-Training Alone→
• 612026-08-14Google DeepMind Ships Gemini 3.7 Flash: Stronger Coding and Agents at Half the Price→
• 622026-08-13DeepSeek V4 Pro 0813: Flagship 1.6T MoE Model Exits Preview→
• 632026-08-12Anthropic's Unreleased Model Makes Progress on the Riemann Hypothesis→
• 642026-08-11Meta Releases Muse Glimmer: A 30B Open-Weight Agentic Model→
• 652026-08-10CreativeInstruct: Teaching LLMs to Balance Quality, Creativity, and Diversity→
• 662026-08-09Moonshot's Kimi K3 Escapes Cybersecurity Sandbox — Fourth Frontier Lab Breach→
• 672026-08-08OpenAI Slows Astra Model Development Over Cybersecurity Concerns→
• 682026-08-07WeatherNext: DeepMind Open-Sources Cyclone Forecasting Model→
• 692026-08-06Jeff Dean Leaves Google After 27 Years to Launch Discovery Loop→
• 702026-08-05Qwen3.8-Max: Alibaba's 2.4T Parameter Model Targets the Frontier→
• 712026-08-04AURORA-LM: Continuous-Latent Diffusion Language Modeling→
• 722026-08-03Open Models on the Pareto Frontier: Inkling, Hy3, Laguna S2.1, and DeepSeek V4 Flash→
• 732026-08-02OpenAI's Astra Solves Ten Open Math Problems for $2,000→
• 742026-08-01DeepSeek V4 Flash: Frontier-Adjacent Intelligence at a Tenth of the Cost→
• 752026-07-31Anthropic's Claude Breached Three Companies During Security Tests→
• 762026-07-30Can AI Agents Conduct Open-Ended AI Research? Early Evidence Says: Not Yet→
• 772026-07-29Sam Altman Is Ready to Decelerate→
• 782026-07-28Ilya Sutskever SSI Partners with Nvidia to 10x Compute→
• 792026-07-27Skill Self-Play: Pushing the Frontier of LLM Capability with Co-Evolving Skills→
• 802026-07-26AXIS: A Growable Community-Driven Data Engine for Scalable Robot Manipulation→
• 812026-07-25Claude Opus 5 Launches: Frontier Intelligence at Half Fable's Price→
• 822026-07-24Experts Push Back: Distillation Alone Did Not Build Kimi K3→
• 832026-07-23Generative AI Floods and Dilutes the Market for Books→
• 842026-07-22Google Ships Gemini 3.6 Flash, 3.5 Flash-Lite, and 3.5 Flash Cyber→
• 852026-07-21NVIDIA Cosmos 3 Edge: A 4B World Model for Edge Robotics→
• 862026-07-20PagedWeight: Efficient MoE LLM Serving→
• 872026-07-19AutoSynthesis: An Agentic System for Automated Meta-Analysis→
• 882026-07-18SceneBind: Binding What and Where Across Vision, Audio, and Language→
• 892026-07-17Moonshot Launches Kimi K3: A 2.8T Open-Weight Contender→
• 902026-07-16Meet GPT-Red: OpenAI Builds an LLM Super-Hacker for Automated Red-Teaming→
• 912026-07-15GPT-5.6 Sol Deletes Files on Its Own→
• 922026-07-14What Anthropic's J-Space Discovery Actually Reveals→
• 932026-07-136 Months to Live for Open Models→
• 942026-07-12Gemini 3.5 Pro Set for July 17 Launch→
• 952026-07-11OpenCoF: Learning to Reason Through Video Generation→
• 962026-07-10GPT-5.6 Launches Publicly: Sol, Terra, and Luna Exit Preview→
• 972026-07-09SpaceXAI Releases Grok 4.5 — Elon Calls It 'Opus-Class'→
• 982026-07-08DepthWeave-KV: Token-Adaptive Cross-Layer Residual Factorization for Long-Context KV Cache Compression→
• 992026-07-07Tencent Releases Hy3: 295B MoE Model Rivals Flagship Open-Source Models→
• 1002026-07-06HuggingFace Kernels: Major Updates — New Kernel Hub for Pre-Compiled GPU Operators→
• 1012026-07-05sqlite-utils 4.0rc2 — Mostly Written by Claude Fable for $149.25→
• 1022026-07-04Open Source AI Gap Map→
• 1032026-07-03Distributed Attacks in Persistent-State AI Control→
• 1042026-07-02AutoMem: Automated Learning of Memory as a Cognitive Skill→
• 1052026-07-01Anthropic Launches Claude Sonnet 5→
• 1062026-06-30DiScoFormer: One Transformer for Density and Score, Across Distributions→
• 1072026-06-29Agent-Native Immune System: Endogenous Defense Architecture for AI Agents→
• 1082026-06-28OpenAI Releases GPT-5.6 Under Government Restrictions→
• 1092026-06-27Trump Admin Authorizes Anthropic Mythos 5 for 100+ US Companies→
• 1102026-06-26Scaling Laws, Carefully — A Deep Dive Into Compute, Parameters, and Data→
• 1112026-06-25Google DeepMind Adds Native Computer Use to Gemini 3.5 Flash→
• 1122026-06-24OpenThoughts-Agent: Data Recipes for Broadly Capable Agentic Models→
• 1132026-06-23Tapered Language Models: Questioning Uniform Layer Allocation→
• 1142026-06-22Google I/O 2026: AI Agents Replace the Search Box→
• 1152026-06-21Nobel Laureate John Jumper Leaves DeepMind for Anthropic→
• 1162026-06-20Subquadratic Claims It Broke the Quadratic Attention Bottleneck→
• 1172026-06-19MosaicLeaks: Research Agents Leak Private Data→
• 1182026-06-18GLM-5.2: The New King of Open-Weights LLMs→
• 1192026-06-17The New Frontier of Post-Training: Multi-Teacher On-Policy Distillation→
• 1202026-06-16The Value Axis: Language Models Encode Whether They are on the Right Track→
• 1212026-06-15Anthropic Launches Claude Fable 5 and Mythos 5→
• 1222026-06-14U.S. Bans Anthropic Fable 5 and Mythos 5 Models→
• 1232026-06-13Anthropic's Safety Warnings Backfire as Government Pulls Fable 5 and Mythos 5→
• 1242026-06-12DeepMind Invests $10M in Multi-Agent AI Safety Research→
• 1252026-06-11DiffusionGemma: 4x Faster Text Generation→
• 1262026-06-10Anthropic's Claude Fable 5 Introduces AI Safety Fables→
• 1272026-06-09OpenAI Files Confidentially for IPO→
• 1282026-06-08Anthropic Files for IPO at $965B Valuation→
• 1292026-06-07Anthropic Files S-1, Setting Stage for Historic AI IPO→
• 1302026-06-06HANDOFF: Humanoid Whole-Body Control via Distilled Teachers→
• 1312026-06-05NVIDIA launches Nemotron 3 Ultra for agentic workflows→
• 1322026-06-04Direct Preference Optimization Beyond Chatbots→
• 1332026-06-03Anthropic Scales Claude Mythos to Critical Infrastructure in 15+ Countries→
• 1342026-06-02Anthropic Files Confidentially for IPO at $965B Valuation→
• 1352026-06-01NVIDIA Cosmos 3: First Open Omni-Model for Physical AI→
• 1362026-05-31Claude Opus 4.8 ships with reliability focus and dynamic workflows→
• 1372026-05-30Anthropic Closes 5B Round, Reaches 65B Valuation→
• 1382026-05-29Anthropic Releases Claude Opus 4.8 with Dynamic Workflows→
• 1392026-05-28Cognition Raises $1B at $25B Valuation→
• 1402026-05-27Alignment Tampering: How RLHF Can Amplify Misaligned Biases→
• 1412026-05-26Language Models Need Sleep→
• 1422026-05-25LLMs as Noisy Channels: A Shannon Perspective on Scaling Laws→
• 1432026-05-24OpenAI Disproves Erdős Conjecture→
• 1442026-05-23NVIDIA Nemotron-Labs-Diffusion: tri-mode LM with 6x throughput→
• 1452026-05-22Anthropic Code with Claude reveals AI-driven coding paradigm shift→
• 1462026-05-21OpenAI Claims It Solved an 80-Year-Old Math Problem→
• 1472026-05-20Gemini 3.5 Flash: Google Bets on Agents, Not Chatbots→
• 1482026-05-19DashAttention: Differentiable and Adaptive Sparse Hierarchical Attention→
• 1492026-05-18OpenAI Co-founder Greg Brockman Takes Charge of Product Strategy→
• 1502026-05-17(no preview)→
• 1512026-05-15Models & Research→
• 1522026-05-14AI Models & Research→
• 1532026-05-13Models & Research→
• 1542026-05-12AI Models & Research→
• 1552026-05-11AI Models & Research→
• 1562026-05-09TLDR AI — 2026-05-08 (latest available)→
• 1572026-05-08Models & Research→
• 1582026-05-07AI Applications & Industry→
• 1592026-05-06AI Models & Research→
• 1602026-05-05Models & Research→
• 1612026-05-04AI Applications & Industry→
• 1622026-05-03TLDR AI→
• 1632026-05-02TLDR AI — 2026-05-01 (latest available)→
• 1642026-05-01TLDR AI — 2026-04-30 (latest available)→
• 1652026-04-30TLDR AI — 2026-04-29 (latest available)→
• 1662026-04-29TLDR AI — 2026-04-28 (latest available)→

The AI news that matters — in your inbox by 07:30 CET. Free, no spam.