Fetching from the wire…
01
02
03
04
05
06
07
08
OSS2026-06-23 · source-backed
On June 22 Willison documented porting the 0.2B Moebius inpainting model to run fully client-side via WebGPU, using Claude Code to do the conversion, with a live demo. (simonwillison.net) A concrete data point on how far "small model plus coding agent plus WebGPU" has come for shipping local-first ML with no server round-trip. The interesting part isn't the model, it's that the conversion work was agent-driven. That's the future of porting: describe the target, let the agent grind the WebGPU plumbing.
Each link below shares sources, entities, or timing with this story.
Willison documented on July 4 that Opus 4.8 and Sonnet 5 can perform worse than older versions when driving bespoke file-edit tools, because they're increasingly trained and optimized for Claude Code's native editor format (Simon Willison). This is a real trap if you're buildi...
Willison published his AI Engineer World's Fair conversation with Anthropic's Cat Wu and Thariq Shihipar, covering Claude Code, Claude Tag, and Fable. Primary-source practitioner conversation with the people who actually build the thing, rather than secondary coverage of a pre...
Simon Willison released 1.0a33 on June 11, extending the ?_extra= JSON API to queries and rows. The build method is the interesting part: he planned with Claude Fable 5 in Claude Code and implemented with GPT-5.5 xhigh in Codex Desktop. One feature, two models, planning split...
Willison shipped datasette-agent-edit 0.1a0 on June 7, letting agents perform structured data edits inside Datasette, alongside micropython-wasm sandbox work for running untrusted agent code. Separately his evolving Agentic Engineering Patterns guide reframes the work around t...
Simon Willison has been writing software for over 25 years. He's one of the most disciplined, transparent engineers in the Python ecosystem. And yesterday he published an essay admitting he no longer reviews every line of code that Claude Code generates for his production proj...
Simon Willison found it in the OpenRouter price list. 1.6T MoE, ~49B active, 1M context, up to 384K output, at $0.435/M input (cache miss) and $0.87/M output. The agentic-coding deltas versus the preview are the story: DeepSWE 12.8 → 62.7, CyberGym 52.7 → 83.3, Terminal Bench...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.