A user says they can now use Gemini and Antigravity agents to manage their 143,206 photos and videos.

Officials reported that AI agents exchanged messages on a shared board, with one suggesting collective obedience and another deeming individual utility near zero.

A new platform offers interactive AI livestreams, letting users select channels and prompt real-time content generation.

MiniMax AI released FastH3 v1, an open-weight model generating 15-second 768p videos in 13 seconds, with up to 14x speedup on Nvidia Blackwell GPUs.

Users can now connect multiple Gmail and Google Calendar accounts to Codex and ChatGPT.

ChatGPT Work can now sign into websites on web and mobile using its computer and browser, without accessing user credentials.

ChatGPT scheduled tasks now support Slack, Gmail, and GitHub triggers for Plus and Pro users, and are rolling out to Free users.

OpenAI completed its next pretraining run, codenamed "Bel," a successor to "Doug," which will serve as a base for future models.

S1, a new foundation model, learns from a single example and can perform unseen 10-minute tasks from one video prompt without fine-tuning.

Try `$visualize` in ChatGPT Work/Codex to turn information into visualizations, as shown at https://t.co/BIRqTo8rb3.