Simon Willison runs a 27-minute GPT-6 Astra agent task to generate OSM-based running routes
Simon Willison · rss · 2026-09-13
Simon Willison used ChatGPT Work with GPT-6 Astra (Max) to generate 5K and 10K looping running routes from his home using OpenStreetMap data. The agent worked for 27 minutes and delivered exactly what he asked: an embedded visualization plus downloadable GPX and GeoJSON files.
- The model's self-described workflow: Nominatim for geocoding, Overpass to download local OSM roads and trails, then local loop calculation.
- Rendering used the visualize skill, producing an HTML file with embedded D3 sourced from allow-listed CDNs; the CSP only permits cdnjs.cloudflare.com, esm.sh, cdn.jsdelivr.net, unpkg.com and a few font origins—others fail silently.
- Route geometry lives in a <script type="application/"> element.
- His main criticism: the actual code the agent ran was not visible in the ChatGPT UI, a transparency anti-feature.
More from coding & agent
- MiniMax Design plugs into Blender via MCP to drive AI video from 3D references — Hailuo_AI · 2026-09-14
- A TypeScript-to-Rust migration revives the classic Rewrite It In Rust meme — DanielLockyer · 2026-09-14
- gemini-cli PR fixes nested .gitignore trailing-slash patterns wrongly anchored — dylanyunlon · 2026-09-14
- Building a 'race your own ghosts' game with Gemini's Astra agent — LexicalLegend · 2026-09-14
- ChainForge's evalstats brings statistical rigor to LLM eval visualization after six months of work — IanArawjo · 2026-09-14
- RiftPoint: a Rust-backed multiversal checkpointer for LangGraph agents — goldenphoenix713 · 2026-09-14