GPT-5.6 Sol, running in Codex at a low thinking level, completed a peaceful Factorio game in 13.5 hours of in-game time. The experiment used a Python-based game manager that let the model control the game programmatically while preserving constraints faced by human players, including limited reach and fog of war. The run was recorded as a screencast, and the manager also tracked production statistics alongside research milestones. The author had planned to stop after blue science, but Sol performed well enough to continue toward launching a rocket.
The run showed both progress and limitations. Sol handled earlier stages better than expected, and markedly improved on prior GPT-5.5 attempts, which had failed to use transport belts. But difficulties scaling production emerged later and slowed progress. Sol struggled to connect some factory systems, ultimately compensating by repeatedly carrying supplies between areas to keep rocket production going. The author sees Factorio as a useful, visually accessible way to study model capabilities and build intuition about how models handle complex tasks, while noting that programming offers a related comparison. Future tests are planned with other models, different thinking levels, and games that include enemy attacks.
Summary generated by AI from the linked article. hn.today is not affiliated with Hacker News or Y Combinator.
Anthropic has restricted Claude from helping with video game modding, decompiling, and reverse-engineering, amid growing use of AI to alter games. The limits are broad enough to disrupt legitimate development and basic troubleshooting, and Anthropic has not publicly explained the restrictions. (mashable.com)
OpenAI’s GPT-6.1 Terra is a midrange model focused on visual reasoning, computer use, and technical communication, with faster computer-use performance than Astra and Sol. It costs $1 per million input tokens and $5 per million output tokens, positioning it between the cheaper Luna and pricier Sol. (simonwillison.net)
Pope Leo XIV argued that AI’s future depends on human choices and that systems should serve people, not be treated as persons. He called for human oversight of high-stakes decisions, steps to ban lethal autonomous weapons, and fair access to AI’s benefits. (garymarcus.substack.com)
Researcher Jiahao Shao says Anthropic blocked or downgraded his access after flagging his work as training a competitor. He says Claude only helped run jobs and analyze results, not generate training data. (twitter.com)
A language model was post-trained on short Dwitter demos to generate plausible continuations of unfinished code and explore alternative visual effects. The project treats demo remixes as a kind of programming grammar, where tiny code changes can produce strikingly different results. (datavorous.com)
A study of knowledge workers found that generative AI changes critical thinking by shifting effort from producing answers toward checking and integrating AI output. Workers who trusted AI more reported using less critical thinking, while confidence in their own skills was linked to greater effort. (dl.acm.org)
Today's best Hacker News stories, summarized and screenshotted, one email a day.