Anthropic published Learning more about Claude's mathematical capabilities on August 10, filed under the slug riemann-zeta — placing it in the growing genre of frontier labs putting their models against open or hard problems in pure mathematics rather than benchmark suites.
It landed hard on Hacker News: 169 points and 117 comments in about ten hours, a comment-to-point ratio that usually signals argument rather than agreement. Claims about model-assisted mathematics tend to draw immediate scrutiny over what the model actually contributed versus what a human collaborator supplied or verified — worth reading the discussion alongside the post itself before taking the framing at face value.
Sources: Anthropic, HN discussion