Rubbish Check
BBC Business · 8 September 2026
source“OpenAI says it cracked 90-year-old maths problem in 88 hours”
R4/ 10
Selective
Rubbish Rating — 1 = base fact, 10 = pure rubbish
12345678910
In short
Rubbish Talk rates BBC's headline that OpenAI "cracked" a 90-year-old maths problem in 88 hours a 4/10 because the company only resolved two of the four statements the Millennium Prize demands, and the result is unverified and disputed by a rival mathematician, neither of which the headline signals.
The Verdict
Selective. The headline correctly hedges with "OpenAI says", which is honest sourcing, but "cracked" implies a completed solve when OpenAI itself resolved only half the required proof and explicitly said it is not claiming the Millennium Prize. The genuinely newsworthy complication, a rival mathematician's public accusation that his team's unpublished work reached OpenAI before its own effort began, is real news buried deep in the article and entirely absent from the headline.
What actually happened
OpenAI says an internal, unreleased AI model directed roughly 10,000 AI agents that deployed up to 10,000 AI agents in parallel over approximately 88 hours, with computing costs running into the millions of dollars, producing a claimed solution to the Navier-Stokes existence and smoothness problem. The proof was produced by an internal AI model, not human mathematicians, and only one Millennium Prize problem has ever been solved before. NYU mathematician Tristan Buckmaster says he and an Anthropic researcher were independently working the same problem and that information about their progress reached OpenAI before its announcement.
Key facts
- 88 hours: time OpenAI's ~10,000-agent swarm took to reach its result, per the article and independently reported by other outlets.
- Only 2 of 4: OpenAI's solution resolved two out of the four statements the Clay Institute's formulation requires; the article states OpenAI is not claiming the $1m prize.
- The proof reportedly used around 10,000 AI agents over 88 hours, followed by 17 hours of Lean formalisation, and the proof was produced by an internal AI model, not human mathematicians, meaning it is not yet peer-reviewed by mathematicians outside OpenAI.
- $10m: the article's own estimate of the compute cost, based on 3 million exchanged messages and 130 billion output tokens.
- Disputed priority: Buckmaster says he learned "information about our progress had been passed to OpenAI" days before the announcement.
What to watch for
Watch whether the Clay Mathematics Institute or independent mathematicians formally verify the proof, since OpenAI's result is self-reported and self-graded so far. Also watch whether Buckmaster's priority dispute escalates, since OpenAI's rebuttal ("we cannot rule out that de-identified data…helped improve our models") leaves the door open rather than closing it.
About this scoreThe R-Score is Rubbish Talk's editorial opinion on how far a headline's framing sits from what the underlying facts support. It is a judgement about presentation and emphasis, not an allegation that any outlet has acted dishonestly. Every figure we rely on is linked under Receipts so you can check it yourself.
