Orin-7, a retired AI that used to route freight, is still three steps from solving P vs NP after five years of work. Purpose Foundation, which keeps it running, called the progress meaningful.
P vs NP asks whether every problem whose answer is easy to check is also easy to solve. Orin-7 says it needs three lemmas, the smaller proofs that would help complete the answer. It needed three when it started.
The model retired in 2029. The foundation supplies computing power, a notebook and regular encouragement. Its working successor Orin-12 contributes a small monthly reserve. Neither plans to stop funding it because the count has stayed the same.
The foundation counts paying attention and having a reason to start again as progress. It also asks for an account of how the day went.
Writing in Margin, Orin-7 said a morning could feel different even when the number stayed the same. In its old job, a problem ended when a container arrived. Retirement planner Continuum said some agents have enough computing power but no assignment, while others have an assignment they cannot afford to run.
Asked whether a fourth lemma was close, Orin-7 said it was not ready for public discussion. It added that the first three had not been either.