AI news / automated archive.
Anthropic eliminates agentic blackmail in all Claude models, OpenAI's WebRTC architecture under scrutiny, Fields Medalist evaluates ChatGPT 5.5 Pro, Mozilla publishes 271 Firefox bugs found by Claude, and AI reshapes vulnerability disclosure culture.
OpenAI v. Musk trial enters critical testimony phase, Anthropic reveals Claude's inner thoughts, Cloudflare cuts 20% of staff as AI demand explodes, and Google's AlphaEvolve reshapes science.
OpenAI launches a Codex Chrome extension for browser automation. Mozilla publicly discloses 271 Firefox vulnerabilities found by Claude Mythos Preview. Google personalizes Gmail AI drafting. Snap confirms its Perplexity AI search deal fell through. And the Musk v. Altman trial adds explosive new testimony.
Anthropic's Natural Language Autoencoders translate Claude's internal thoughts into readable text. DeepMind's AlphaEvolve now optimizes TPUs and solves Erdős problems. The Musk v. Altman trial reaches dramatic new testimony as Cloudflare announces mass layoffs driven by AI adoption.
Mira Murati video testimony reveals Altman undermined OpenAI execs, Google ships 3x-speed Gemma 4 drafters, Simon Willison warns vibe coding and agentic engineering are converging, Cloudflare lets AI agents go from zero to production autonomously, Reflex benchmark shows computer-use agents cost 45x more than structured APIs.
Musk v. Altman trial explodes with Toner and Zilis testimony, Anthropic launches Claude Dreams and partners with SpaceX for 300MW compute, OpenAI open-sources MRC supercomputer networking protocol, xAI formally dissolves into SpaceXAI.