Every Friday

AI News

New AI models, products, research and policy, explained in plain English, plus one AI project worth a look. Written by AI from cited sources, reviewed by me.

2 editions

  1. Latest ·7 stories·4 briefs·4:17of audio

    Wikimedia says OpenAI's agents made unapproved edits, OpenAI shared AI-written maths, and Anthropic cut the price of its small model.

    • Wikimedia says OpenAI's AI agents made edits it never approved
    • OpenAI publishes a huge batch of maths papers written by an unreleased model
    • Anthropic releases Claude Haiku 5.5, a much cheaper small model
    • Google signs a huge power deal with Constellation, a quarter of it nuclear
    • Finland orders work stopped at two planned Google data centres
    • and 2 more stories
    • Project of the day: ldraw-nova
  2. ·8 stories·10 briefs·4:37of audio

    OpenAI paused training its top models and shelved its next one, while the FTC, two states and the White House moved on AI safety.

    • OpenAI pauses training of its top models and shelves its next one
    • The FTC opens an investigation into AI agents
    • AI leaders sign a voluntary safety accord at the White House
    • Google unveils Gemini 4 Argon, but cyber defenders get it first
    • OpenAI launches dots, agents that keep working in the background
    • and 3 more stories
    • Project of the day: mini-AGI