HN Reads the Erdős Breakthrough as Human-Guided AI Math, Not Magic

The headline was irresistible, but HN refused to read it as a simple AI solved math moment. Scientific American reported that 23-year-old Liam Price, working with ChatGPT Pro, produced a solution to a 60-year-old Erdős problem about primitive sets. The article argues that this case matters because the route appears genuinely new, and because mathematicians who reviewed it think the idea may have uses beyond this one problem.

That last part is what gave the story weight. The article quotes Terence Tao saying the problem may have been easier than expected once the first move was seen correctly, and Stanford mathematician Jared Lichtman says the proof's underlying connection could travel further. This is a stronger claim than the usual AI-math story, which often collapses into either benchmark theater or a rediscovery of tools experts already know.

HN still kept its distance from the hype. One commenter linked the chat and said the raw proof output was not what impressed them; the interesting part was whether a language model could propose a route humans had not taken. Another highlighted a quote from the article that the raw output was actually poor and had to be sifted by an expert. That framed the result less as autonomous theorem proving and more as a hybrid workflow: machine proposes, human interprets, experts verify.

Several commenters also noted that Erdős problems cover a wide range of difficulty, so the label alone should not be treated as a universal score for mathematical intelligence. HN's stronger test is narrower and more practical: did the model contribute a non-obvious idea that survives expert scrutiny? In this case, the answer looks more interesting than in many previous rounds of AI-math hype, precisely because the community can point to a novel angle instead of just a flashy headline.

So the mood was skeptical, but not dismissive. HN is clearly willing to credit AI when it helps uncover a fresh line of attack. It just does not want the human role erased from the story. The source article is Scientific American, and the discussion is in the HN thread.