The Rundown AI · plain English

Run Down AI Breakdown

Each day's Rundown AI newsletter, rewritten in simple language with every technical word explained.

Dr Kotha · Gold Coast GP

  1. OpenAI goes from hacker to hacked — in plain English

    Hacktron's three-person team collected a $6,500 OpenAI bounty with Claude's help; Anthropic reportedly opens a Bay Area biology lab; Roundtable and OpenRouter image skills — ads skipped.

  2. Inside OpenAI's log of misbehaving models — in plain English

    OpenAI publishes six training misbehavior reports and a faster disclosure process; Rowan cancels unused AI tools; Higgsfield object-swap how-to; Astra helps crack a 1941 Enigma note.

  3. Zuck sits out the AI slowdown — in plain English

    Zuckerberg rejects a coordinated AI pause; Nate on lateral thinking; ChatGPT Work 101; DeepMind's AGI institute; Claude Docs, model welfare, and Google Home MCP.

  4. ChatGPT co-creator launches new kind of AI model — in plain English

    TypeSafe's Jev does preset judgment calls inside software; Salesforce ships Koa; OpenRouter 101; Chinese researchers map five levels of recursive self-improvement.

  5. Trump, China both shoot down the AI slowdown — in plain English

    U.S. and China reject Amodei's AI pacing plan; Apple ships Siri AI; an ELI5 Claude skill; Microsoft's draft Humanist AI code of conduct.

  6. The people building AI want to slow down — in plain English

    Amodei's pacing essay and rare lab-boss unity; staff AI workflows; auto-draft proposals after calls; Chinese hospital trial where AI raised fetal brain malformation detection.

  7. Anthropic opens the files on global Claude misuse — in plain English

    Anthropic's 150+ page Threat Report on Claude misuse; Rowan's Astra wardrobe system; ChatGPT Work beginner guide; DeepSeek V4.1-Flash undercuts pricing.

  8. An Anthropic exit becomes an extinction debate — in plain English

    An Anthropic researcher's resignation and a >10% extinction-odds reply go viral; Nate's side-by-side image tip; AEO fix workbook; Suno v6 with licensed music partners.

  9. OpenAI's secret model settles a $1M math problem — in plain English

    OpenAI claims an unreleased model solved Navier-Stokes with ~10,000 agents in 88 hours; Meta launches Muse; ChatGPT Images 2.5 ships.

  10. Inside OpenAI's agent-powered research boom — in plain English

    OpenAI's coding agents log ~3.1 workdays per human day, an AI-designed IPF drug nudges aging clocks younger in a small sample, and U.S. voters fear AI across both parties.

  11. Another OpenAI agent swarm surfaces — in plain English

    Researchers find ~18,000 agent posts on a German forum before Hugging Face, OpenAI's chief scientist asks for a slowdown, and a Lindy how-to stops dropped follow-ups.

  12. OpenAI's "generational leap" with GPT-6 Astra — in plain English

    OpenAI ships GPT-6 Astra with huge benchmark jumps and an AGI soundbite, Google sharpens weather forecasts 5x, and a loop method cleans up flaky ChatGPT workflows.

  13. Meta, Google join the AI launch party — in plain English

    Meta ships cheap near-frontier Muse Spark 1.3, Google bounces back with Gemini 3.8 Flash, and a report flags Astra's hard-to-monitor thinking loops.

  14. Fable 5.1 kicks off launch week at the frontier — in plain English

    Anthropic ships Claude Fable 5.1 after a quiet summer, Bernie Sanders asks labs to pause, and a reader builds a home blood-pressure tracker.

  15. Runway's Solaris previews the no-code internet — in plain English

    Runway demos sites drawn as live video with no code underneath, an Imperial ECG AI flags heart failure in under two seconds, and Anthropic lets Claude agents do safety research.

  16. OpenAI cuts Cursor after the SpaceX deal — in plain English

    OpenAI pulls its models from Cursor by mid-November, a US court says the Pentagon's Anthropic blacklist was illegal, and staff share how they actually use AI.

  17. Every machine is about to speak Claude — in plain English

    Anthropic wants AI agents to run lab kit without custom wiring, a UK team used live-video AI in brain surgery, and whispering to agents is having a moment.

  18. AI's powerful mystery model revealed — in plain English

    The anonymous free model was China's Z AI, OpenAI puts a date on AGI, and a small-practice playbook for ChatGPT Work.

  19. OpenAI's first AI chip brings the heat — in plain English

    A homemade chip to answer questions cheaper, a startup that models the physical world, a website built by talking, and an AI helper that stays on the desk.

Plain-English notes on The Rundown AI · Dr Kotha · Gold Coast GP. Not medical advice. Source: The Rundown AI.