Will AI solve one of 129 major mathematical conjectures before year X?

Prediction market on manifold. The definition of major conjecture for the purposes of this market is one of the 129 "open problems" in this Wikipedia article as of 1 August 2025: https://en.wikipedia.org/wiki/List_of_conjectures If a valid proof found by an AI is posted, then all years after the posting date will resolve YES. The proof has to be accepted by a majority of the mathematical community (this market will stay open for a significant amount of time after a claimed proof to allow for disputes to arise). It does not count if the problem was solved by a human beforehand. If the proof was a human-AI collaboration, that is likely not sufficient for a YES-resolution, unless an overwhelming amount of the work (>90%) was done by the AI. If the AI produces an essentially complete proof, but humans reformulate the proof into easier-to-read prose, that would be sufficient. Traditional AI like SAT-solvers/SMT-solvers count as AIs for the purposes of this market, but any problem translation or tinkering with these tools count as human work (if performed by humans). I will likely not count problems that only involve such tools, since there is usually still a lot of human work needed for such proofs (e.g. proofs of the empty hexagon number or the pythagorean triple coloring problem). A few weeks after the start of a year, if I haven't found a proof claim and none has been posted in the comments, that year will resolve NO. (Edit Mar 2026) If a problem has multiple parts or multiple versions, then all parts or the strongest version has to be proven for the purposes of this market. (Edit Mar 2026) If the proof essentially uses a human-written article, preprint, blog post or similar that is less than 3 months old, I might categorize that as a human-AI collaboration at my disgression. Citing any older works will not count as a collaboration. I will not bet on this market

24h Volume: $314.438. Liquidity: $1,500. Resolves: 1/1/2050.