Story · Simon Willison
Generating running routes with GPT-6 Astra and ChatGPT Work (Simon Willison)
blog post · Story page
Simon Willison had GPT-6 Astra in ChatGPT Work build 5K and 10K running loops from OpenStreetMap data in 27 minutes, then could not retrieve the Python it ran, apparently because the thread had been compacted. His ask: keep the pre-compaction text and expose it to the agent through a tool call.
In plain words
- Simon Willison had an artificial intelligence assistant plan running routes that started and ended at his home.
- It calculated the routes using local roads and trails from OpenStreetMap, a source of map data.
- He later could not retrieve the computer instructions it ran, apparently because the system had shortened the conversation.
- He argues that saving earlier conversation text would help users recover details about how the assistant completed their tasks.
Appeared in
- An agent deleted an AML control, and benchmark scaffolds do the model's work
Sep 14, 2026 · in the sections
Subscribe
Get the brief in your inbox
Pick daily, weekly, or both. Nothing is gated either way: every issue is on the site and in the feeds.
- Weekdays at 8:45am IST, one lead story and 6 to 9 items.
- Sundays, an argued synthesis rather than a recap.
- One click to leave, and quiet days say so in the subject line.