GPT-5.6 Sol, running in Codex with low thinking level, completed a peaceful Factorio run in 13.5 hours of game time. The run used a Python-based game manager to allow programmatic interaction while enforcing human-like restrictions such as player reach and fog of war. A full timelapse is available on YouTube.
Sol exceeded expectations in some areas but showed clear weaknesses, particularly in late-game production scaling. The rocket launch was achieved only through significant manual effort, as Sol struggled to connect production systems and relied on ferrying supplies between factory areas. Despite this, the run marked a clear improvement over previous attempts with GPT-5.5, which failed to use transport belts at all.
The author plans to release a detailed commentary video and test other models and thinking levels, including non-peaceful runs.
Source: Hacker News · Summarized by HeadlinesBriefing