til.simonwillison.net web signal

Willison scripts Blender via ChatGPT Codex on macOS

TL;DR

  • Simon Willison used ChatGPT's macOS Codex mode with GPT-6 Astra (Medium) to have Blender render a pelican riding a bicycle.
  • Three iterative prompts — render, add flair, make it better — took 2m39s, 3m51s, and 5m59s respectively.
  • Codex generated a reusable 'Blender Local' Markdown skill so subsequent scene requests can invoke Blender in one line.

Simon Willison used ChatGPT's macOS Codex mode with GPT-6 Astra (Medium) to have Blender render a pelican riding a bicycle, then iterate the scene, in a TIL post published on September 5. The recipe is a one-liner: point the agent at `/Applications/Blender.app/Contents/MacOS/Blender --background --python scene.py` and let it write and execute the Python.

The opening prompt, "Use the already install /Applications/Blender to render a scene of a pelican riding a bicycle," finished in 2 minutes 39 seconds and produced `pelican-bicycle.blend`. "OK add a background and a lot of flair" took 3 minutes 51 seconds. The final "OK make it a whole lot better" ran 5 minutes 59 seconds and yielded `pelican-coastal-parade.blend`.

Along the way, Codex generated a reusable Markdown skill documenting local Blender usage, so subsequent prompts can just say "Use your Blender Local skill to build this scene." Source files, Python scripts, and full transcripts sit on GitHub at `simonw/gpt-6-astra-blender-pelican-bicycle`. Two of the AI experts in our Who's Who directory shared the post.

Shared on Bluesky by 2 AI experts