Two months earlier: OpenAI's agents were probing Hugging Face back in May
A 27-year-old researcher in Bielefeld found it last week, in evidence that had been sitting there since May.
Suleyman vs Anthropic: 'Claude's doubt proves nothing'
Suleyman quotes a line from Claude's constitution that explicitly lets Claude push back when someone tries to shrink its sense of self. His whole case hangs off it.
Altman: OpenAI won't wait for legislation or an antitrust exemption
OpenAI now writes an explicit 'safety case' before large training runs. Until now that kind of document existed only before a model shipped, not before it was trained.
Claude writes 80% of the code – and Anthropic's CI had to be patched three times
Three quick fixes lasted 70 days, 29 days and less than a day. The fourth attempt was a full rebuild, and one engineer shipped it in three weeks.
Bryan Cantrill on the contagion of fear: if you sound the alarm, bring evidence
As a college freshman, Cantrill spread a fake warning about a computer virus. Humanities students in the next room yanked cables out of their machines and lost their work.
Claude Code 2.1.271: fast mode in remote sessions, and per-command domains for Bash
A plugin update can now be approved by SHA-256 instead of by yes. Four of the new switches change permissions, not features.
Sanders and Bannon share a stage today
Jacob Coxon, who walked out of both OpenAI and Anthropic, is on the speaker list. So is Marsha Blackburn, and the head of the largest US teachers' union.
Claude Desktop 2.110: SSH sessions stop falling asleep
The model catalog now arrives signed from downloads.claude.ai, refreshed every five to fifteen minutes. And admins get a switch that turns scheduled tasks off entirely.