AI Signal 185
Tao warns AI effort may 'flatten' open math problems before researchers reach full potential
↗
Illustration only Photo by Monisha Selvakumar on Unsplash
In a quote reposted by Simon Willison, mathematician Terence Tao argues that AI is enabling the open-problem pool of mathematics to be depleted rapidly enough to discourage researchers from sharing promising directions.
For engineers building AI research assistants or automated theorem provers, the warning frames capability as a cost: the faster an AI can solve a stated problem, the less incentive anyone has to publish the problem publicly in the first place. Tao is not attacking AI tools but describing an externality they impose on the problem-generating pipeline that those tools themselves depend on. The post carried here is a short quotation collected by Willison on 9 September 2026, so the underlying Tao essay is not reproduced and the full argument cannot be checked.
Written by elseif from the cluster below · every claim links back to a sourceThe three things worth knowing
Simon Willison reposted a Terence Tao passage on 9 September 2026 in which Tao describes open mathematical problems as a non-renewable resource that is now being 'mined' at speed.
Tao writes that 'even the rumor of someone working on a problem can trigger a massive amount of AI-powered effort to flatten it before the original research project has time to reach its full potential.'
Tao concludes that the resulting incentives may push researchers toward not sharing promising research directions with the broader community, which he says would 'reverse centuries of traditions of open science and do serious long-term damage to the future of the field.'
THE READ
What the cluster adds up to.
This item is a single quotation collected and reposted by Simon Willison on 9 September 2026; no other feed in the material carried it. The substance is a short excerpt from mathematician Terence Tao warning about how AI-assisted research changes the incentive to publish open problems. Because the post is a quotation, only Tao's words are in the material; the larger essay they were taken from is not, so the argument has to be reconstructed from the quoted passage alone.
Tao's central framing is that good open problems are a finite stock being 'mined in a non-renewable fashion.' Once a problem is published, AI tooling can apply large-scale effort to close it, and the population of unsolved problems that is worth publishing therefore shrinks faster than it would have under human-only research. For an engineer, this recasts AI capability not as a neutral productivity multiplier but as something that can exhaust the resource it consumes, which is the pool of publicly stated open problems.
The mechanism Tao names is competitive rather than purely individual. He observes that 'even the rumor of someone working on a problem can trigger a massive amount of AI-powered effort to flatten it before the original research project has time to reach its full potential.' The implicit race condition is that the marginal value of a problem to its originator drops sharply once the problem becomes known, because other actors can apply more compute to it than the original researcher can match.
From that race condition Tao draws an incentive conclusion: researchers may stop sharing 'promising research directions with the broader community,' which he says would 'reverse centuries of traditions of open science and do serious long-term damage to the future of the field.' The concrete consequence is a change in publication norms rather than in tooling, and the cost would fall on whoever depends on the public problem pool, including the AI systems being trained or benchmarked against it.
The material does not give us the longer Tao essay this passage comes from, any proposed mitigation, or any response from other mathematicians. A working engineer should read this as one researcher naming a structural concern about the problem-publishing pipeline, not as evidence that secrecy has already displaced open sharing in mathematics; the note itself offers no data on publication behavior, only Tao's warning about where the incentives may now point.
Written by elseif from the cluster below · checked for specifics the sources never containedTHE CLUSTER