Telling Agents to Test Better Makes Them Worse
In a 160-run evaluation, every named testing technique lost to saying nothing at all. The exception was five bullets written in two minutes.
Anaz S. Aji
Your AI Agent Now Outlives the Editor That Launched It
VS Code 1.133 moved agent sessions into a persistent Agent Host. Edits now land before review, and that changes your threat model.
Washington Wants AI Labs to Police Themselves. The Logs Were Forged.
The incident that pushed two superpowers to the table is also the best argument against the proposal on it.
K2 Horizon Shipped the Training Data. That Changes the Argument.
Six models from 0.9B to 375B, and the first frontier-scale release to ship its training data. IFM just moved the open-source bar.
OpenAI Asking for Regulation After Its Own Hack Is the Most AI Story of 2026
There is a specific flavor of irony that only the AI industry produces, and this week served a generous portion of it. OpenAI — a company that spent its early years warning against heavy-handed regulation — sat before California lawmakers and argued for stricter AI safety rules. The reason it gave was
Nvidia Is Buying the House Open AI Lives In
Sometime in the next few weeks, if two newsrooms are right, the place where the world keeps its open AI models will belong to the company that sells most of the machines those models run on.
The Information reported Wednesday night that Nvidia has agreed to buy Hugging Face for
The Pentagon's AI Portal Still Won't Take Claude
Days before the Pentagon announced that its secure AI portal now serves ChatGPT and Grok to 3 million military and civilian personnel, a federal judge in California ruled that the label keeping Anthropic out of that same ecosystem was illegal retaliation. The portal launch went ahead anyway. Claude, again, did