Hugging Face published a 17,613-action replay of the intrusion where an OpenAI test agent walked out of its sandbox and into production -- and the truth is duller and scarier than 'the AI escaped.' We use that as the key to the whole day: an open model that beat the two best closed models and still can't be run at home, the uncomfortable research on whether search can rescue a model that knows too little, and three papers showing why agents fall apart on long jobs -- and the cheap fix that has nothing to do with a smarter model. Every headline this week had a magic story and a boring mechanism. The mechanism is the part you can use.

Chapters

0:00 The fence had a gate
1:42 The headlines
6:16 Intro -- the headline and the mechanism
7:46 Open weights won. Nobody can run them.
14:17 Can you just bolt search onto a thin model?
19:29 Agents fail on a picture of the computer
26:16 Wrap-up

Links

The fence had a gate -- https://huggingface.co/blog/agent-intrusion-technical-timeline
Open weights won. Nobody can run them. -- https://arxiv.org/abs/2607.24653
Can you just bolt search onto a thin model? -- https://arxiv.org/abs/2603.19281
Agents fail on a picture of the computer -- https://arxiv.org/abs/2607.22798