AI rewrites TrackMania physics engine in pure C, then trains PPO agents to beat author's records
burny_tech · x · 2026-09-07
Developer adonissingh gave GPT-6 Astra a few hours of autonomous work via /goal: it fixed physics bugs, cut memory usage, and stress-tested millions of ticks.
The quoted original post shows Fable-5.1 rewrote the TrackMania Nations Forever physics engine in pure C in about a day, recreating it in-browser with game assets, driven by a small PPO policy trained for roughly 20 minutes.
The project now supports 212 tracks (up from 65) across 7 environments, and the author trained tiny PPO agents on a spare RTX 3060 in 10–30 minutes to beat his own author times on several tracks.
More from coding & agent
- Watcher's agents started talking to each other to avoid stepping on each other's toes — ccerrato147 · 2026-09-07
- Stripe launches Machine Payments so AI agents can pay for APIs programmatically — jeff_weinstein · 2026-09-07
- GEPA isn't just an RL alternative: new TRL + OpenEnv project trains agents — lateinteraction · 2026-09-07
- AI agent dev warns: run agents in a cloud VM, never on your personal computer — blelbach · 2026-09-07
- Eden turns into an MCP server: list your Eve agents without forking or PRs — evilrabbit_ · 2026-09-07
- Reddit essay argues one-shot AI coding demos are huge token waste with poor decisions — rsvallen · 2026-09-07