AI Weekly Malaysia

Back to items Summaries

Using Blender with coding agents on macOS

ID
21769
Status
summarized
Published
05 Sep 2026, 11:51 PM
Fetched
07 Sep 2026, 9:06 AM
Provider
Simon Willison
Category
developer-ai
Original URL
https://simonwillison.net/2026/Sep/5/blender-coding-agents-macos/
Source URL
https://simonwillison.net/atom/everything/

Summary

Score
6.0
Created
07 Sep 2026, 9:06 AM
Tags
Audience
developersvibe_codersai_agent_users

What happened

Simon Willison reports that frontier models have become very good at driving Blender via its Python API. Using ChatGPT Codex on macOS, he prompted the model to render scenes (e.g., a pelican riding a bicycle), iteratively refine them, and even produce movies by rendering image sequences combined with ffmpeg. The setup is simple: install the full Blender Mac app from blender.org and prompt the coding agent to use it.

Why it matters

If you build or experiment with coding agents, this is a concrete, low-friction way to test how well frontier models control a complex desktop application through its Python API—useful for evaluating agent capabilities beyond text and code generation. Try it yourself to gauge where model-driven 3D generation currently stands.

Discussion angle

How reliable are frontier coding agents at driving non-trivial desktop software like Blender through APIs, and what does this tell us about agent readiness for real-world tool-use beyond code editing?

Top