AI news for builders and product teamsUpdated Nov 18, 2026, 11:00 UTC
AI news, straight from the source.
Releases, research, and ideas for developers and product teams. Updated every hour.
Across the feeds
Newest first
Episode 251 of the Last Week in AI podcast covers Anthropic redeploying Claude Fable 5 and launching Claude Sonnet 5, new Google tools, and business and research updates including Etched, Baidu, Agility Robotics, DeepSeek and China's LongCat-2.0.

Last Week in AI released its 250th episode, recorded 06/27/2026 and hosted by Andrey Kurenkov and Jeremie Harris. Topics included US government gating of frontier AI, OpenAI's GPT-5.6 Sol rollout, Anthropic's Mythos, compute supply chain competition, and GLM 5.2.

Episode 249 of the Last Week in AI podcast, recorded 06/17/2026, covers Anthropic cutting off access to Fable 5 and Mythos 5 after a US government order tied to alleged jailbreaks, and SpaceX's IPO at a roughly $1.75T valuation followed by a move to acquire Cursor for $60B.

Last Week in AI released episode 248, recorded 06/12/2026, covering Anthropic's Claude Fable 5 release, Apple's Siri AI announcement at WWDC, and a range of business, open-source, policy and safety news items.

Episode 247 of the Last Week in AI podcast, recorded June 3, 2026, covers Anthropic's Claude Opus 4.8 release, Microsoft's Scout assistant and MAI models, Anthropic's $65B Series H and IPO filing, and MiniMax-M3 among other AI news items.

Moonshot AI released Kimi K3, a 2.8T-parameter mixture-of-experts model that ranks #2 on the Vals AI index and #3 on Artificial Analysis's Intelligence Index, with weights scheduled for July 27. The release narrows the open-to-closed performance gap to a claimed 3-5 months.

The UK AI Security Institute found the cybersecurity capability gap between open-weight and closed frontier models has narrowed, with GLM-5.2 and DeepSeek V4-Pro performing like closed models released four to seven months earlier. Kimi also announced Kimi K3, a 2.8 trillion parameter model, while Demis Hassabis proposed a FINRA-style standards body for frontier AI testing.

Together AI and Y Combinator announced a partnership to launch the first dedicated YC GPU cluster, giving YC portfolio startups access to compute for inference and training with short-term sprints at long-term rates instead of long-term commitments. The cluster is running at full utilization today.

Ahead of AI's new article explains how reasoning models get multiple reasoning-effort settings, using OpenAI's recently released GPT-5.6 family as the motivating example. GPT-5.6 ships in three sizes, each with roughly five or six effort levels ranging from Light to Ultra.

Together AI published an explainer describing what 99%, 99.9% and 99.99% inference uptime tiers require architecturally, including the failure domains each must survive, and listed questions to ask inference providers before committing.

Thinking Machines Lab released Inkling, a multimodal mixture-of-experts model for text, image, and audio reasoning. Together AI is offering day zero access to Inkling on its inference platform, including serverless availability with a 1M context window and OpenAI-compatible APIs.

Together AI announced updates to its GPU Clusters covering passive health checks, auto node repair, a rebuilt Slurm-on-Kubernetes stack (Slinky 1.0), external OIDC for Kubernetes RBAC, startup scripts, a new cluster details view, and an acceptance-test opt-out.

Interconnects writer warns open-weight AI models could face U.S. restrictions within roughly six months, citing reported White House discussions and distillation debates led by Anthropic. The piece argues there are no official details and that proposed limits would mainly affect Chinese-origin models and government use.

Amazon and the University of Michigan developed HydroShear, a physics-based simulator that models tactile shear forces so robots can learn contact-rich manipulation policies in simulation that transfer to real hardware. It achieved a 93% average success rate across four tasks on a Franka robot with GelSight Mini sensors, versus 34% for TacSL and 58-61% for FOTS.

Amazon Science released Turnstile, a Rust proxy that sits between an agent harness and a model backend to record exact token IDs, log probabilities, loss masks, and weight-version boundaries during generation for reinforcement learning. Validations included a text-only coding agent and a multimodal computer-use agent, whose harnesses were left unchanged.

Mistral announced that its Studio product now offers a system of record for AI prompts and skills, providing versioning, ownership, traceability, rollback, classification labels and audit logs. The feature is available today to Mistral Studio customers.

Ollama announced it has raised $88M from Benchmark, Theory Ventures, 8VC, Y Combinator and angel investors. The company says it serves 8.9 million developers and is used by 85% of the Fortune 500.

Mistral AI has introduced Robostral Navigate, an 8B model for embodied robot navigation that uses only a single RGB camera and no depth sensors or LiDAR. It reports 76.6% success on the R2R-CE validation unseen benchmark, outperforming multi-sensor approaches.

Import AI 464 covers an AI-written GPU megakernel on KernelBench-Mega, rising AI automation on the Remote Labor Index, the OSWorld 2.0 computer-use benchmark, and JD's Oxygen AI Item Center.

Lilian Weng published a post on harness engineering for recursive self-improvement, covering harness design patterns, context engineering methods such as ACE and MCE, automated workflow search approaches including ADAS and AFlow, and Meta-Harness. The post also traces the concept of recursive self-improvement to I. J. Good (1965) and Yudkowsky (2008).