What math should AI tackle? FAR contributor says choosing will matter more

A contributor-endorsed summary describes FAR as a pipeline that starts with a research direction, searches the literature for open problems, attempts them at scale and filters results for mathematicians. According to the summary, its combinatorics pilot narrowed 51,110 papers to 4,717 apparently open, attemptable conjectures. It produced 1,050 claimed new resolutions; 598 passed automated judging and 77 were recommended for expert review. The summary says the authors manually checked 15 selected outputs and found all mathematically correct, though one had already been solved elsewhere. A contributor to the work said deciding what AI should try to solve would become increasingly important as models improve.