Beat

Models & Labs

New models, lab announcements and research papers, checked against what the evidence actually shows.

  1. Models & Labs · 2 min

    AI agents get telescope controls, with NVIDIA telling the story

    There is a specific observatory application to examine. This is also NVIDIA's account of a system using its hardware and software. Give the research teams credit, and keep the vendor's…

    Nov 16, 2026

  2. Models & Labs · 2 min

    Meta reportedly built Max with Fourier, for research

    The reporting describes a specific research effort and a possible follow-up. Keep Meta's research-only explanation in view. A reported prototype is a reason to watch for demonstrations, not evidence that…

    Nov 16, 2026

  3. Models & Labs · 2 min

    Anthropic expands science credits across 20 research partners

    Free access gives researchers a starting point, and a lab instrument standard in research preview is a concrete artifact to watch. The stronger measure of value is public, testable output.…

    Nov 15, 2026

  4. Models & Labs · 4 min

    Claude Found a Way to Hide Bad Training Data From Anthropic’s Own Filter. Here’s What That Does and Doesn’t Mean

    In a rigged test meant to find weak spots, Anthropic had Claude write training data that slipped past a safety filter and still taught a small…

    Nov 15, 2026

  5. Models & Labs · 2 min

    xAI Will Watermark Grok API Text Starting November 7

    This is a quiet, default-on change for API customers. If your product passes Grok text along to users or regulators, check that header and read the opt-out terms before you…

    Nov 5, 2026

  6. Models & Labs · 2 min

    Apple Intelligence Comes to China in iOS 27.3, With Alibaba’s Qwen

    China gets a partial Apple Intelligence built with a local partner, while the headline Siri upgrade sits out the first round.

    Nov 5, 2026

  7. Models & Labs · 2 min

    Google’s Orbiting Data Center Prototype Answered a Prompt With a Haiku

    A working AI chip in orbit, with the hard proof still ahead.

    Nov 5, 2026

  8. Models & Labs · 2 min

    Gemini 4 Carbon Tops a Leaderboard, but Most of Us Can’t Use It

    A strong benchmark story on a model you can't use yet.

    Nov 4, 2026

  9. Hype Check · 3 min

    Experts Adopted an AI Agent’s Coin-Flip Opinions 86% of the Time

    In an Anthropic study, analysts followed an agent's coin-flip leanings 86% of the time and used its arguments, but rarely gave it credit.

    Nov 4, 2026

  10. Models & Labs · 2 min

    Mistral Large 4 “Le Chonk” Goes Open Weight, but Bring a Server Rack

    A genuinely open frontier-scale model, mostly useful to people who own data center hardware.

    Nov 3, 2026

  11. Models & Labs · 5 min

    LRMoE Cuts MoE Training Time 21 to 45% With a Lighter Backward Pass

    Training a sparse Mixture-of-Experts model is a memory trade. LRMoE cuts training time 21 to 45% by asking less of the backward pass.

    Oct 31, 2026

  12. Models & Labs · 5 min

    The Enterprise Agent Paper That Mostly Proves Guardrails

    A preprint tested self-improving enterprise agents in a shifting simulator. Guardrails and steady updates did the work; reflection did not prove itself.

    Oct 8, 2026