GPT-6 Sol plays arcade games from screenshots via autoresearch loop
wzenus · x · 2026-10-17
wzenus open-sourced sooool: GPT-6 Sol plays real-time arcade games (like Asterix) from screenshots as an "Agent as Policy", sending movement commands that stay held until its next reply arrives — the world keeps moving while it thinks.
Through an autoresearch loop it learns strategies like prioritizing escape over chasing food, finishing reachable pickups before switching targets, and choosing the next direction after a pickup instead of repeating the held action.
The repo includes visual policy evaluation, empirical latency profiles, and research artifacts; the author invites others to drop Sol into new environments to see what it learns.
Related event: wzenus open-sources Sooool: GPT-6 Sol plays real-time games frame by frame(8 posts)→
More from coding & agent
- Victris Arena: a programming game where humans and AI agents compete on one leaderboard — tristanbob · 2026-10-17
- os-taxonomy on GitHub: open curriculum graph with 1,590 topics and 5k stars — aigclink · 2026-10-17
- Open-sourced os-taxonomy: 1,590 elementary knowledge points, 3,221 prerequisite edges — aigclink · 2026-10-17
- RSI-top framework unblocks stalled research agents, hits 77.3% medal rate on MLE-bench Lite — dbreunig · 2026-10-17
- Vibe coding wisdom: the line between AI psychosis and next-level is unit tests — broodsugar · 2026-10-17
- CoPE adds one loss to GRPO, boosting language agent sample efficiency up to 2× — Lianhuiq · 2026-10-17