Generating running routes with GPT-6 Astra and ChatGPT Work
- ID
- 23916
- Status
- summarized
- Published
- 13 Sep 2026, 7:56 AM
- Fetched
- 13 Sep 2026, 9:02 AM
- Provider
- Simon Willison
- Category
- developer-ai
- Original URL
- https://simonwillison.net/2026/Sep/12/astra-running-routes/
- Source URL
- https://simonwillison.net/atom/everything/
Summary
- Score
- 6.5
- Created
- 13 Sep 2026, 9:02 AM
- Tags
- Audience
- developersai_agent_usersai_ml_learners
What happened
Simon Willison used ChatGPT Work with GPT-6 Astra (Max) to generate 5K and 10K running routes from his address using OpenStreetMap data via Nominatim and Overpass. The agent ran for 27 minutes and produced an embedded D3 visualization plus downloadable GPX and GeoJSON files, but the actual code it executed was invisible in the UI and unrecoverable after thread compaction.
Why it matters
If you build or use agent systems with context compaction, you should ensure pre-compaction text is preserved and retrievable via tool calls—Willison lost the entire code his agent ran because compaction discarded it and the model couldn't reconstruct it. This is a concrete transparency failure to design around in your own agent architectures.
Discussion angle
When building agent systems that compact context, what's the minimum audit trail you need to preserve—code, tool calls, intermediate outputs—and how do you expose it without overwhelming the user?