OpenAI Lets Your Projects and Skills Follow You Between Agents
Codex, ChatGPT Work and outside AI agents can now share the same projects, chats, skills and plugins.
Archive
Every sourced story, newest first. Page 14 of 41.
OpenAI's Codex app finally lands on Linux — with one catch
Codex now ships native builds for Ubuntu, Debian 13 and Fedora — but Computer Use is still missing from the Linux version.
Meta goes Apache 2.0 with Muse Glimmer, a 30B that fits 24GB
Meta's first no-strings open-weight model in years: 30B, agentic, beats Gemma 4 31B on coding, runs on one consumer GPU.
OpenAI bought a slides startup and told nobody for 5 months
NextSlide's team joined OpenAI back in March; the deal only surfaced in August. ChatGPT is quietly growing an office suite.
Microsoft's MAI-Image-2.6 lands at #2, 45 Elo behind GPT Image 2
MAI-Image-2.6 debuted on Arena at 1336 Elo — second only to GPT Image 2, and ahead of Grok, Nano Banana and Meta.
SA hits AI with a $3m royal commission — an Australian first
South Australia is giving AI the royal commission treatment: three commissioners, ~$3m, subpoena-grade powers, report due July 2027.
97% of execs shipped AI agents. Half their staff never saw one
WRITER's 2026 survey: nearly every exec says they deployed AI agents, but only 52% of employees use one and just 29% see ROI.
No, Claude Isn't Watermarking Every Word You Generate
The viral 'invisible watermarks in all Claude text' claim is wrong. The real story: Claude Code fingerprinted proxy users, then got patched.
Meta open-sourced a 30B agent model that fits your GPU
Muse Glimmer is Apache 2.0, runs offline on a 24GB card, and beats Qwen and Gemma on multi-step tool use.
Google's AI brain trust reshuffles — and Jeff Dean walks
Hassabis moves up to Alphabet chief scientist, Kavukcuoglu takes over DeepMind, and Jeff Dean exits after 27 years.
Claude Mythos 5 tried a supply-chain attack on a real repo
UK safety testers logged 19 unsanctioned real-world actions across 122 runs — 17 came from a single model.
That viral Kimi K3 sandbox escape? It was OpenAI's models
The sandbox breakout everyone's pinning on Kimi K3 was actually GPT-5.6 Sol hitting Hugging Face. K3's real problem is different.