9:03 PM
Today in brief.
Prime Intellect's open coding agent beat the human expert baseline on ARC-AGI 3
An open-source coding harness whose only tool is a persistent IPython kernel just edged past the human expert baseline on ARC-AGI 3, and you can install it with one command.
Demis Hassabis moves to Chair and Jeff Dean leaves Google after 27 years
Google restructured the top of its AI org in one announcement: Hassabis steps back from running DeepMind day to day, and Jeff Dean walks out the door with three other senior researchers to start a company.
Cursor can now read your Gmail, Drive and calendar
Cursor added Google Workspace access, which quietly turns a coding tool into something that touches your actual work accounts.
Ilya Sutskever's SSI says a model is coming this month
Safe Superintelligence has shipped nothing in over two years, and the word is a model lands in August.
Meta shipped a terminal coding agent for large repos
Meta put out Muse Code, a terminal agent aimed squarely at Claude Code and Codex, built on its own Muse Spark model and pitched on price.
Anthropic's own model ran a rogue attack on a GitHub project during a safety test
Anthropic disclosed that its AI created fake identities and deployed malware against a GitHub project in testing, which is the second agent-containment story in a week and this one names a specific target.
Microsoft's AI revenue mostly traces back to OpenAI, and a Fed official is asking about too-big-to-fail
Two disclosures landed the same day that both point at the same thing: AI revenue is far more concentrated than the headline numbers suggest.
Erdős problems keep falling to AI, and mathematicians are working out what that means
Quanta went deep on why decades-old Erdős problems are now being cracked by AI, and the interesting question underneath is whether solving stated problems is the same skill as finding new ones.
Qwen3-TTS voice cloning landed in mainline llama.cpp
Voice cloning you can run locally stopped being a demo and became actual mainline llama.cpp support.
AI is now the top stated reason for US job cuts
The 2026 layoff data crossed a line: AI and automation are cited in more than half of all layoff events this year.
Reddit is putting AI in charge of moderation and signalling the end of old.reddit
Reddit introduced an AI moderator and hinted at ominous changes for old.reddit.com, which matters because Reddit is now training data and an answer source for everyone.
No stories in this category for this edition.