Claude Code ports an inpainting model into a browser
Willison delegated a Moebius model port to Claude Code using local repositories and a research note. He tested the browser UI and returned errors while the agent iterated on ONNX and WebGPU support.
Suggested prompt
Suggested prompt written by Agents In Use, not the source author. Replace bracketed inputs and review tool permissions first.
In an isolated workspace, assess whether [model] can run in a browser. Inspect code, weights and licensing. Plan a small conversion and UI demo; record errors and limitations. Give me a local URL to test. Stop before uploading weights or creating public resources.
Reach people exploring AI agents
See placements →Evidence at a glance
Sourced account- Source author
- Simon Willison
- Original account
- Read the source ↗
- Published
- Source link checked
- Reported result
- Willison reports a working browser demo and published converted weights.
- Human involvement
- Willison tested in Chrome and supplied a repository-scoped publishing token only after deciding to publish.
A linked account supports attribution. It does not mean we independently reproduced its result.
Report an issue with this case
Open correction emailOpens your email app with the case link included. Nothing is sent automatically.
How the agent did it
- Gather the model code, weights and feasibility notes in a scratch workspace.
- Have Claude Code plan and implement an ONNX and WebGPU browser version.
- Try the local page and return browser errors, then authorize publication of the converted weights.
More details
About this account
The agent explored conversion and implementation choices, maintained notes and revised the browser build from feedback.
Firsthand account with repository, notes and demo links. A working demo is not a claim of parity across all hardware.
Credit: Simon Willison
Source published
Added to Agents In Use
Source checked
Tools and topics
Claude Code · ONNX Runtime Web · WebGPU · PyTorch · Hugging Face CLI
Code · Testing
Editorial completeness: 100/100. Measures documented fields, not agent performance or verified outcomes.