What the claim says

The claim reads, in English rendering: "AI solved a decades-old math problem … 10,000 stepped up for a 'relay solve.'" Two assertions carry the weight. First, that AI contributed to resolving a problem that had stood open for decades. Second, that the work proceeded in relay fashion, with 10,000 units taking part. The Korean counter used in the headline — "gae," a general-purpose counter — leaves the object unstated, which becomes relevant to the final rating. The citation attached to the claim pointed to a rules page for a formal mathematics prize framework, signaling that the "decades-old math problem" at issue sits within an established prize context rather than informal puzzle culture.

How it was checked

The address cited in the claim, ending in "e-rules/," returned a 404 error when reviewers attempted to load it. Rather than dropping the citation, the team rerouted through the site's /millennium-problems/rules/ path — the canonical rules page for the Millennium Prize Problems — and secured the intended content there, confirming the framework the claim invoked is genuine and current. In total, 12 primary sources were reviewed before the determination was recorded.

Where it lands

The core survives the paper trail: the prize-rules framework is verifiable despite the broken link, and the relay framing matches the claim as published rather than overstating it. Two compression issues temper the score. The headline's "10,000" arrives without its object, forcing readers to infer whether the count refers to compute units, model instances, or problem sets; and "solved" reads broader than the underlying record, which supports a more measured account of AI's contribution to a problem that predates modern computing. Neither issue rises to distortion — both are matters of headline shorthand. On that balance, the claim rates Mostly True.