GPT-5.6 Sol completes peaceful Factorio run in 13.5 hours

Production statistics were recorded throughout the session and plotted against science research milestones. The author had planned to stop after the blue‑science stage, but the model’s unexpected effectiveness led to a continuation. Scaling issues appeared later, slowing progress, and the final rocket launch required significant manual effort from the model to ferry supplies and patch late‑game production gaps. Compared with earlier attempts using gpt-5.5, which failed to use transport belts, the new run shows clear improvement.

The experiment is presented as a step toward autonomous Factorio play as a way to study model behavior, noting the similarity to programming while being more observable for casual viewers. The author plans to test other models, thinking levels, and non‑peaceful runs in the future.