Which Millennium Prize Problem will AI solve next?
BSD being the next Millennium Prize Problem solved by AI looks possible but not the leading candidate. The current evidence points more strongly to other problems drawing the nearest verified attention, so I would price Yes below the market.
Analysis
The market is asking a very specific sequencing question: not whether Birch and Swinnerton-Dyer will eventually be solved by AI, but whether it will be the first qualifying Millennium Prize Problem announced as solved by an eligible AI lab before the end of 2027. That makes the bar much higher than a simple long-run forecast. The recent context does not show a confirmed public BSD breakthrough, only scattered claims, commentary, and speculation that AI systems may have touched BSD-adjacent ideas or that a lab may be exploring several Millennium problems in parallel. Under the market rules, that is not enough unless an eligible lab explicitly announces a complete solution, and the current evidence falls well short of that threshold.
The strongest signal in the recent information is that the attention around AI mathematics has been concentrated elsewhere, especially on Hodge conjecture claims and on broader, still-contested math progress narratives involving Navier-Stokes. Even if those reports are partly speculative, they matter for this market because they indicate where labs and the public seem to be focusing their verification effort and publicity. BSD is a famously deep number theory problem, and while it is not obviously less prestigious than the others, it does not appear to be the most publicly advanced or most visibly promoted target at the moment. In a race where only the first qualifying announcement counts, being merely one of several possible targets is a disadvantage.
There is still a nontrivial case for Yes because AI labs have strong incentives to make headline-grabbing mathematical announcements, and BSD is a natural candidate for a major breakthrough if a model or hybrid workflow manages to produce a convincing proof. The existence of a large, well-capitalized set of eligible labs increases the chance that one of them is working on exactly this kind of result in secret. However, the market price already reflects some of that optionality, and the lack of any concrete, verified BSD announcement by late 2026 means the base rate is still low. For Yes to cash, BSD must not only be solved, but also beat every other qualifying problem to the announcement line, which is a difficult sequencing constraint that materially reduces the probability relative to the chance of a future AI solution in the abstract.
Arguments
For
- Arguments for Yes: BSD is a canonical, high-value target that a leading AI lab could plausibly choose for a landmark announcement.
- Arguments for Yes: The abundance of well-funded eligible labs raises the odds that at least one team is pursuing deep theorem-proving work on BSD.
Against
- Arguments against Yes: There is no confirmed public evidence that BSD has been solved by AI, only speculative references to possible progress.
- Arguments against Yes: Other Millennium problems appear to have more visible momentum, making BSD less likely to be the first announced success.
Key drivers
- The market resolves only if BSD is the first qualifying problem publicly announced as solved by an eligible AI lab.
- Recent attention appears to be concentrated more on Hodge and Navier-Stokes than on BSD.
- No verified public announcement of a BSD solution has emerged from an eligible lab.
- Any qualifying announcement must be complete and explicit, not a partial result or progress claim.
Risk factors
- A lab could quietly develop a BSD proof and announce it before rivals announce other Millennium solutions.
- A surprise breakthrough from a major frontier lab could jump ahead of the more visible candidates.
- News flow may lag internal progress, causing the market to underestimate near-term announcement risk.
- If multiple problems are solved around the same time, alphabetical tie-breaking could change the resolution outcome.
Scenarios
Best case
An eligible AI lab unexpectedly announces a complete BSD solution before any other Millennium problem is publicly solved, and the announcement is clear enough to qualify immediately.
Most likely
AI labs continue making noisy but inconclusive mathematical progress, with any eventual qualifying announcement more likely to concern a different problem than BSD.
Worst case
No eligible AI lab announces a qualifying BSD solution by the deadline, or another Millennium problem is announced first and BSD loses the race.
More from this day
- PoliticsKalshi23d
Will 3 mg THC beverages be federally legal to sell?
AI9%MKT92%Edge-83HypedA 3 mg THC beverage is very unlikely to remain federally legal to sell on Nov. 16, 2026 under the newly reported hemp rules. I put the Yes probability at 9%, with the market appearing heavily overpricing legalization persistence.
- economyPolymarketEnded
What will the median home value in the SF Metro area be on September 30?
AI16%MKT62%Edge-46HypedThe evidence strongly suggests the San Francisco metro median home value will remain above $1,176,000 on September 30, so the chance of a Yes outcome is low. The market price looks too optimistic for Yes given the recent price level and momentum.
- SportsKalshi37d
When will Bill Belichick leave North Carolina?
AI34%MKT69%Edge-35HypedI make Belichick’s chance of being out before Dec. 1, 2026 materially lower than the market suggests. The guaranteed contract and reported administrative support make an immediate exit possible but not the base case.