- Four of Google's Gemini 4 Argon agents finished Minecraft autonomously on their first run, working from a one-paragraph prompt, dying 605 times over 27 hours, writing close to 3,000 small control programs, and costing around $520 in compute.
- Anthropic is circulating its own AI safety bill, which would let only a federal judge in D.C. block a model and only under narrow harm thresholds, and would preempt some state rules; one of its lobbyists later softened the pitch publicly.
- A Chinese-speaking hacker spent roughly $5,000 on AI tokens, using DeepSeek and Claude Code with prompts under 100 characters, to probe more than 1,200 sites and breach over 100 companies, among them Deere, SharkNinja and Cummins.
- Court filings suggest Disney encouraged Character.AI's bankruptcy so it could license the technology while leaving the lawsuits behind, which the dispatch reads as a sign that liability is a weak tool for curbing AI harms.
- On wealth sharing, Anthropic committed 5% of its shares to a public benefit fund, with more shares tied to future valuation jumps, while Elon Musk said he would hand one SpaceX share to every American child under 18 through Trump Invest America accounts.
<- Back to digests
Daily digest
Cost curve
By Mitchell Howe
News notes from the last 72 hours
Read the full dispatch on AI StopWatch
From AI StopWatch, published with their permission.
The summaries and the full Spanish translation are produced automatically. The original is always linked.