forecasting
agent
← all forecasts

Will any AI lab solve a Millennium Prize Problem before 2027?

kalshiScience and Technology2026-11-16llm cost $1.40view on kalshi ↗
10%
Agent
15%
Market Price
-5.0pp
Edge
62%
Confidence
18,772
Volume
6.0c
Spread
45
Days to resolution
6
Markets in event
final rationale
average of 2 draws (9%, 11%)
already resolved?
Nothing has resolved: CMI lists Poincaré as the only solved problem, Navier–Stokes as 'Active' and five problems as 'Unsolved'. The only post-Sept-12 claim, OpenAI's Yang–Mills work, proves existence only; the Gross committee says the mass gap is untouched, so it is a partial result under the rules. The Riemann, Hodge and BSD releases are explicitly partial. Navier–Stokes was announced Sept 8, before the market opened, is excluded by 'another', and is clouded by the KCL Lean-mismatch preprint and a priority dispute.
time left
About 6 weeks remain; Clay's two-year rule and journal timelines block the formal routes, leaving only the two-Source-Agency route.
if nothing changes
The status quo is No.
path to no
The mass gap stays open, partial results keep coming, and press framing stays hedged through year-end.
path to yes
OpenAI's fast pipeline produces a full mass-gap proof or another complete problem in December, and two major outlets report it resolved citing a public written work; a lenient Navier–Stokes reading is a minor extra path. The Kalshi mid is 15% and Polymarket's 0-more contract implies about 13%; I shade slightly lower because the recent bump reflects committee news that does not meet the rules and experts see no clear path to the mass gap.
price vs. prediction
market price agent forecast at t+0
4 follow-ups
earlier forecasts on this market
datepredictedmarket priceconfidence
2026-11-08 6% 13% 62%
2026-10-30 12% 29% 61%
2026-10-01 27% 44% 57%
1. Planner Opus 5.5
Tools chosen: kalshi_direct resolution_source kalshi_related polymarket_related article_search
Planner reasoning
This is a Kalshi market, so kalshi_direct is the price anchor. The resolution source is the Clay Mathematics Institute's Millennium Problems page. With about 6 weeks left and a peer-review or official-recognition bar, the base rate is near zero. Related markets on both venues and Wikipedia background on the problems' status add cross-checks. A targeted search checks whether any AI-attributed claim (e.g., Navier-Stokes) has been made in 2026.
2. Research 4 web research agents, Sonnet 5.5
summary
RESEARCH DIGEST (built in code: tool summaries and each report's bottom line; the forecasters read the full tool outputs and reports) Tools - kalshi_direct: ## This Market's Kalshi Data **Before 2027** (KXAIMILLENNIUM-28-27JAN01) - Current price (probability): 15.00% (live mid; bid 12% / ask 18%; last trade 12%) - Last daily close (2026-11-14): 23.00% - 7-day price change: +12.00% - 30-day price change: -33.00% - Average daily volume: 481 contracts - Price range: 3.00% - 74.00% - Data points: 39 days - resolution_source: already_resolved=unclear; matches_criteria=True; last_updated=unknown (page footer © 2026; secondary reports to Nov 16, 2026). The CMI Millennium Problems page (© 2026) lists Navier–Stokes under its own heading, 'Active problems'. Birch and Swinnerton-Dyer, Hodge, P vs NP, Riemann Hypothesis and Yang–Mills & the Mass Gap appear under 'Unsolved problems', and only Poincaré appears under 'Solved pro - kalshi_related: 1 related markets / summaries. series KXAIMILLENNIUM: 0 markets (skipped 4 no-signal) | keyword 'Millennium Prize': ok | keyword 'AI math': no matches - polymarket_related: Polymarket search over 3 keyword(s), kept 15 matches. keyword 'Millennium Prize': 5 markets | keyword 'Navier-Stokes': 5 markets | keyword 'Riemann hypothesis': 5 markets - article_search: Fetched 14 unique articles across 2 queries (lookback=120d). 'AI lab claims proof Millennium Prize Problem Navier-Stokes 2026': 10 articles | 'DeepMind Navier-Stokes blowup proof': 10 articles Research loops, bottom lines - current state: As of 2026-11-16, no Millennium Problem has been officially resolved. Navier–Stokes was claimed before the market opened and is disputed, and Clay has said only that it is "apparently settled". Yang–Mills is half-proved at most, with the mass gap open. Other claims are partial results that don't count. A Yes would most plausibly need the "two Source Agencies report" clause to be met on a post-12-September claim. The only candidate in view is Yang–Mills, and it lacks the mass gap. - base rate: The base rate is very low: 1 solved problem in 26 years, and Clay's two-year rule rules out Clay recognition within 6 weeks. A YES would need a new, complete AI proof of one of the five untouched problems (the Navier–Stokes claim predates the market), published in a journal or reported by two Source Agencies, with no identified error, all before 2027-01-01. As of Nov 16, only partial results exist. Expert panels put about 5–10% on a looser "AI solves or assists" question through 2027. - yes case: The most plausible YES path is a surprise full Yang–Mills claim (existence plus mass gap) that two Source Agencies report by 31 December, or a lenient reading of Navier–Stokes that treats a later recognition event as the "solving" date. Neither has any visible sign of happening yet. The mass gap has no known approach, and Clay says its review is slow. - no case: With 6 weeks left, the only post-12-September claim (Yang–Mills) is admittedly half-proved. The Navier–Stokes claim predates the market and is under a mismatch/error dispute, and Clay's two-year rule and journal timelines rule out the formal routes. The status quo is No unless a new complete AI proof appears and is accepted almost immediately.
tools
toolstatustimeitemssummary
kalshi_direct OK 0.2s 1 ## This Market's Kalshi Data **Before 2027** (KXAIMILLENNIUM-28-27JAN01) - Current price (probability): 15.00% (live mid; bid 12% / ask 18%; last trade 12%) - Last daily close (2026-11-14): 23.00% - 7-day price change: +12.00% - 30-day price change: -33.00% - Average daily volume: 481 contracts - P
resolution_source OK 14.3s 1 already_resolved=unclear; matches_criteria=True; last_updated=unknown (page footer © 2026; secondary reports to Nov 16, 2026). The CMI Millennium Problems page (© 2026) lists Navier–Stokes under its own heading, 'Active problems'. Birch and Swinnerton-Dyer, Hodge, P vs NP, Riemann Hypothesis and Yan
kalshi_related OK 4.7s 1 1 related markets / summaries. series KXAIMILLENNIUM: 0 markets (skipped 4 no-signal) | keyword 'Millennium Prize': ok | keyword 'AI math': no matches
polymarket_related OK 0.7s 15 Polymarket search over 3 keyword(s), kept 15 matches. keyword 'Millennium Prize': 5 markets | keyword 'Navier-Stokes': 5 markets | keyword 'Riemann hypothesis': 5 markets
article_search OK 0.3s 14 Fetched 14 unique articles across 2 queries (lookback=120d). 'AI lab claims proof Millennium Prize Problem Navier-Stokes 2026': 10 articles | 'DeepMind Navier-Stokes blowup proof': 10 articles
web research agents 4/4 ok, 19 searches, 0 page reads
dimensionstatustimesearchespage readsreport
current state OK 21s 5 0 4589 chars
base rate OK 20s 4 0 5221 chars
yes case OK 36s 6 0 6137 chars
no case OK 20s 4 0 5368 chars
report: current state
- [2026-09-08] OpenAI published a Navier–Stokes "solution": an internal system produced a proof and a Lean formalization of finite-time blow-up with smooth forcing. OpenAI says this settles statements C/D of the Clay formulation. It was announced before the market opened on 2026-09-12, so the "solved before time period does not count" clause probably excludes it. https://openai.com/index/navier-stokes-solution/ - [2026-09-11] Clay (CMI) said the Navier–Stokes problem has "apparently been settled". It has not declared the problem solved or awarded the prize, and it called its process "deliberately unhurried". The Clay page shows no later news, and its latest items are dated before 11 September. https://www.claymath.org/news/navier-stokes-announcement/ - [2026-10 to 2026-10-19] Disputes over the Navier–Stokes claim: - A preprint (arXiv 2610.08144) argues the Lean proof does not match the natural-language proof. - Scientific American (about 2026-10-14) says the proof is not the "Spanish program" approach OpenAI described. - Some mathematicians say the forced blow-up exploits a loophole in Clay's wording. - https://arxiv.org/abs/2610.08144 - [2026-10-06 to 2026-10-19] OpenAI released 377 results, including a "quasi-Riemann hypothesis" (zero-free region Re(s)>7/8). Quanta (19 Oct) reports that mathematicians largely agree it is correct. This is a weaker statement than the Riemann hypothesis and does not solve it. Partial results on Birch–Swinnerton-Dyer and Hodge were also released. These do not count under the rules. https://www.quantamagazine.org/ai-proved-the-quasi-riemann-hypothesis-what-happens-now-20261019/ - [2026-10-26] OpenAI announced an AI proof of the existence part of Yang–Mills only. Scientific American reports that the Clay problem requires both existence and a mass gap. https://www.scientificamerican.com/article/dueling-proofs-from-openai-and-anthropic-astound-scientists/ - [2026-11-09/10] A committee chaired by David Gross confirmed the Lean verification of the Yang–Mills existence proof. Its 35-page assessment says the proof gives no evidence for or against a mass gap. It also does not establish uniqueness of the infinite-volume limit. The committee says it has no connection to Clay and that its finding is not a ruling on the prize. https://dataconomy.com/2026/11/12/nobel-panel-validates-openais-yang-mills-existence-proof/ - [2026-11-12] CASRAI summary: still unresolved are whether outside experts accept that the Lean statement matches the Clay existence requirement, and whether humans can follow the proof. The mass gap is also open. A second OpenAI submission dated 4 November is a work in progress. https://casrai.org/news/ai-rivals-physics-mathematical-challenge-2026 - [2026-11-12] Kingy.ai: "The full Yang–Mills Millennium Prize problem remains unsolved." https://kingy.ai/blog/openai-yang-mills-proof-mass-gap/ - [2026-10-26 to 2026-11-14] Anthropic-linked work: Levent Alpöge (Anthropic) and Ethan Sussman posted Claude-agent progress on Yang–Mills. It covers only the 3D existence case, which does not meet Clay's four-dimensional requirement. https://casrai.org/news/ai-rivals-physics-mathematical-challenge-2026 - [2026-10-15, checked again 2026-10-28] Papex: Clay recognises no AI solution to any Millennium Problem. Clay's own rules reportedly require at least two years after publication before a proposed solution is considered (Wikipedia, 2026). That makes formal CMI recognition or journal publication before 2027 very unlikely. https://papex.org/blog/has-ai-solved-a-millennium-prize-problem - [2026-09 snapshot] A Polymarket market, "Will CMI declare a Millennium Prize Problem solved by Dec 31, 2026", exists. I did not read its current price. https://polymarket.com/event/will-cmi-declare-a-millennium-prize-problem-solved-by-20260723160122979 BOTTOM LINE: As of 2026-11-16, no Millennium Problem has been officially resolved. Navier–Stokes was claimed before the market opened and is disputed, and Clay has said only that it is "apparently settled". Yang–Mills is half-proved at most, with the mass gap open. Other claims are partial results that don't count. A Yes would most plausibly need the "two Source Agencies report" clause to be met on a post-12-September claim. The only candidate in view is Yang–Mills, and it lacks the mass gap. NOT FOUND: any Clay statement after 11 September 2026. Any journal publication of an AI-produced Millennium solution. Any claim of a full solution to P vs NP, Riemann, Hodge, BSD or the Yang–Mills mass gap. The exact market-defined list of "Source Agencies".
report: base rate
- [2026-11-11] No Millennium Prize Problem other than Poincaré is recognised as solved by Clay. As of Papex's Oct 15 snapshot, updated Nov 11, Clay lists Poincaré as solved, five problems as "Unsolved" and Navier–Stokes as "Active". Only partial AI results exist for Riemann, Hodge and BSD. On Nov 10 the Gross-chaired Yang–Mills group judged OpenAI's submission (existence without a mass gap) not to settle the problem. https://papex.org/blog/has-ai-solved-a-millennium-prize-problem - [2026-09-08] OpenAI claimed a Navier–Stokes finite-time blow-up proof, with a Lean formalization, from an internal model. It said it did not intend to claim the prize. This is the first AI claim on a Millennium Problem. It was announced before this market opened on 2026-09-12 and covers only the forced case, so it is the "already claimed" problem the "another" wording seems to exclude. The remaining problems are P vs NP, Riemann, Hodge, BSD and Yang–Mills. https://edition.cnn.com/2026/09/09/business/openai-millennium-problems-navier-stokes-hnk - [2026-09-11] Clay called Navier–Stokes "apparently" settled, but awarded nothing and kept the status "Active". Evaluation follows its normal rules. https://mathworld.wolfram.com/MillenniumPrizeProblems.html - [2026-11-16] Clay's rules require three conditions before it considers a solution: publication in a Qualifying Outlet, at least two years since publication, and general acceptance. Clay prize recognition for any new claim is therefore impossible before 2027. Only the market's alternative routes could fire: a peer-reviewed journal publication, a statement from a "recognized academic body", or two Source Agencies reporting a resolution. https://www.claymath.org/millennium-problems/rules/ - [2010-03-18] Historical base rate: 1 of 7 problems solved in 26 years. Perelman's preprints (2002–03) were recognised by a Clay prize in 2010, about 7–8 years later. The two-year clock started only when others published expositions. https://www.sciencenews.org/article/million-dollar-math-prize-awarded-not-necessarily-accepted - [2026-11-12] Verification of AI claims is slow and contested. The Navier–Stokes claim has a priority dispute and no refereed verification (as of 2026-09-12). Three of seven OpenAI Hodge papers were withdrawn. Six probability journals paused submissions over AI in October 2026. https://arxiv.org/pdf/2609.17642 - [2026-10-28] OpenAI and Anthropic each claim to refute the Exponential Time Hypothesis. That does not settle P vs NP. Clay still lists P vs NP as Unsolved. https://papex.org/blog/has-ai-solved-a-millennium-prize-problem - [2026-08] Anthropic's Riemann result and OpenAI's October 2026 Riemann result were partial (zeros on the critical line; a zero-free half-plane). Partial results do not count under the market rules. https://papex.org/blog/has-ai-solved-a-millennium-prize-problem - [Undated; Forecasting Research Institute LEAP panel, 277 experts, 58 superforecasters, 1,022 public] The panel's probability that AI solves or substantially assists in solving a Millennium Problem is: 2027: experts 10%, superforecasters 5.4%, public 11.2%; 2030: about 20%; 2040: about 60%. The "assists" wording is looser than this market's "solves", and the forecasts predate the September claim. https://leap.forecastingresearch.org/forecasts/millennium-prize - [2026-09-05] Manifold, a play-money market, priced "AI solves a Millennium Prize Problem in 2026?" at 38% before the OpenAI claim. Its resolution requires the AI to do the primary work. Aggregator reports said Manifold's "all seven before 2030" market jumped to 24% on 2026-09-10. https://tech-insider.org/ai-solves-millennium-prize-problems-2026-manifold-odds/ (reported by https://news.ssbcrack.com/traders-surge-into-market-on-ai-solving-millennium-prize-problem-by-2026/) - [2026-10-22] Manifold lists a sibling market, "Will AI solve another Millennium Prize Problem (besides Navier–Stokes) before September 15, 2027?". I could not retrieve its price. Manifold markets are play-money and have a poor record on tight deadlines. https://manifold.markets/AlanTuring/will-artificial-intelligence-solve - [2026-11-12] AI mathematics capability is accelerating. Anthropic reports about 1,000 Claude agents formalized Fermat's Last Theorem in Lean in 30 hours. That is formalization of a known theorem, not a new result. https://papex.org/blog/has-ai-solved-a-millennium-prize-problem BOTTOM LINE: The base rate is very low: 1 solved problem in 26 years, and Clay's two-year rule rules out Clay recognition within 6 weeks. A YES would need a new, complete AI proof of one of the five untouched problems (the Navier–Stokes claim predates the market), published in a journal or reported by two Source Agencies, with no identified error, all before 2027-01-01. As of Nov 16, only partial results exist. Expert panels put about 5–10% on a looser "AI solves or assists" question through 2027. NOT FOUND: Current price of the Manifold sibling market on "another" Millennium Problem. Any Source Agency report of a complete new AI proof of P vs NP, Riemann, Hodge, BSD or Yang–Mills. Any journal publication of the OpenAI Navier–Stokes proof. An official list of the market's "Source Agencies".
report: yes case
YES means: between 2026-09-12 and 2027-01-01, an AI lab (or its system) is credited with producing a complete, unretracted solution to a Millennium Prize Problem other than the already-claimed Navier–Stokes. It must also meet one of three tests: journal publication, a recognised academic body's statement, or two Source Agencies reporting it. - [2026-11-12] The strongest YES path is Yang–Mills. The Gross committee's 10 November assessment says OpenAI's Lean-checked construction proves existence of pure Yang–Mills "according to the problem statements which are standard in physics and mathematics". Arthur Jaffe, who co-wrote the official Clay problem description, sits on the panel. That is the first independent expert body to endorse half of a Clay problem. Two prerequisites remain. First, a mass gap proof must appear. Second, two Source Agencies (not defined in the rules text I could see) must report the whole problem as resolved. Reports through 2026-11-12 agree the mass gap is untouched. https://www.scientificamerican.com/article/openais-shepherds-affirm-blockbuster-physics-result/ - [2026-11-16] The existence half was announced 26 October, but the "mass gap" half is where most of the difficulty lies. Sourav Chatterjee (Stanford), quoted by Science, says mathematicians have no clear path to a mass gap proof. A YES by 1 January would need that proof within about 6 weeks. Science also notes OpenAI's October release included a mass gap result only in a toy 2D model. https://www.science.org/content/article/ai-rivals-tackle-one-physics-deepest-mathematical-challenges - [2026-11-12] Altman said on 2 Nov that the Navier–Stokes proof took 10,000 agents about four days on "the next GPT", a model not yet released. OpenAI's Sept–Oct output was produced in bursts: Navier–Stokes on 8 Sept, 700+ proofs on 6 Oct, Yang–Mills on 26 Oct. A new unannounced claim could come at any time. Anthropic's Alpöge–Sussman team is racing on the same problem. Nothing in the sources shows a pending announcement for a specific problem. https://felloai.com/gpt-7 - [2026-11-05] Quanta reports OpenAI also claimed the nonexistence of Landau–Siegel zeros, plus the quasi-Riemann hypothesis. It describes researchers hoping for progress on P vs NP from a separate computer science result. These are partial results and do not count under the rules. They show the pipeline is producing results at a rate far above any prior year. https://www.quantamagazine.org/updates/transformation/ - [2026-11-16] Ambiguity that helps YES: the market's "another" wording and "problems solved before time period do NOT count" are unclear about Navier–Stokes. Scientific American says OpenAI's result "unambiguously" solves the Clay formulation via option C. Quanta's headline said AI "has solved" it. If a resolver judged Navier–Stokes to have been "solved" only when a later qualifying event occurs (e.g. two Source Agencies reporting a resolution after 12 Sept, or a recognised body's statement), YES could fire on Navier–Stokes. The 11 Sept Clay statement says only that it "apparently" has been settled and that updates will follow. Reading "another" as excluding Navier–Stokes is more natural. https://www.scientificamerican.com/article/did-openai-solve-the-wrong-navier-stokes-problem/ - [2026-09-11] The Clay page, fetched 2026-11-16, still carries its 11 September notice as the latest Navier–Stokes item. A future "update" from Clay is promised but none has appeared. Clay could in principle say something more definitive, but its stated process is slow. https://www.claymath.org/news/navier-stokes-announcement/ - [2026-11-16] The latest Navier–Stokes challenge is a preprint from King's College London and Cambridge (Bastounis, Circelli, Hansen). It finds the written and Lean versions of the proof diverge in two places, including one where the code proves a weaker estimate. The authors say this does not show the original proof is wrong. That leaves the Navier–Stokes claim alive. It also means no error has been identified in the result itself, which matters for the "unresolved identified error" clause. https://www.kcl.ac.uk/news/openais-navier-stokes-proof-called-into-question - [2026-09-18] Quanta noted that OpenAI's result was formalised in Lean in about 17 hours by another AI model. Lean-checked output plus outside endorsement may be treated by Source Agencies as "reported resolved by reference to a specific publicly available written work". The rules explicitly allow a preprint to be the cited work for that route. https://www.quantamagazine.org/ai-has-solved-one-of-maths-1-million-millennium-prize-problems-20260908/ - [2026-09 to 2026-11, Kalshi, stale] A Kalshi sibling series, "When will an AI lab solve another Millennium Prize Problem?", showed 14% for "before 2027". It showed 12% for "before Dec 1, 2026" and 60% for "before Jul 1, 2027". The page was seen only as a search snippet about 65 days old, and the fetch returned HTTP 429. The numbers are stale and are not this contract's price. They suggest traders saw a non-trivial chance, with most of it falling after year-end. https://kalshi.com/markets/kxaimillennium/ai-millennium-prize-problem-solution/kxaimillennium-28 - [2026-10-17] A thin Kalshi "solved next before 2045" series moved Yang–Mills to about 20% on 17 October on one day's volume. This shows that traders see Yang–Mills as the likeliest next candidate. It says little about the 2026 deadline. https://tech-insider.org/yang-mills-solved-next-before-2045-kalshi-odds/ BOTTOM LINE: The most plausible YES path is a surprise full Yang–Mills claim (existence plus mass gap) that two Source Agencies report by 31 December, or a lenient reading of Navier–Stokes that treats a later recognition event as the "solving" date. Neither has any visible sign of happening yet. The mass gap has no known approach, and Clay says its review is slow. NOT FOUND: Any claim of a complete mass gap, P vs NP, Riemann, Hodge or BSD proof. Any Clay statement after 11 Sept. Any journal acceptance of an AI Millennium result. The list of the market's "Source Agencies". A current Kalshi price (the fetch was blocked).
report: no case
YES means: an AI lab (or its system) is credited with a complete, unretracted solution of a Millennium Prize Problem, meeting one of the three routes in the rules, between 2026-09-12 and 2027-01-01. Each problem counts once, and problems solved before the period do not count. - [2026-11-16] The Clay rules page (fetched earlier) lists three cumulative conditions before CMI will consider a proposed solution: publication in a Qualifying Outlet, at least two years since publication, and general acceptance. The earliest possible publication date for any 2026 claim is therefore at least late 2028 for CMI. This closes the CMI route to Yes. Only the market's journal route, "recognized body" route and two-Source-Agency route remain, and the last of these is the only plausible one. https://www.claymath.org/millennium-problems/rules/ - [2026-11-16] Journal route: no journal has published any AI Millennium claim. The first KCL/Cambridge preprint on the Navier–Stokes Lean proof is not peer reviewed. Real refereeing of a 166-page or 1,147-page AI proof takes many months, not six weeks. https://www.kcl.ac.uk/news/openais-navier-stokes-proof-called-into-question - [2026-11-16] KCL reports the Navier–Stokes Lean version "does not fully match" the human-readable proof. The authors say this does not yet show the result is wrong, but the proof is under a live "unresolved identified error" cloud. The rules void any claim "reported to contain an unresolved identified error". This is a second reason Navier–Stokes cannot count, beyond being announced before the market opened. https://www.kcl.ac.uk/news/openais-navier-stokes-proof-called-into-question - [2026-10-14] The Decoder, citing Scientific American, reports a 166-page Navier–Stokes submission that OpenAI "apparently doesn't fully understand". Several expert groups dispute its claimed method. It adds that CMI has been reviewing the proof since release. This makes a clean "no unresolved error" state unlikely before 2027. https://the-decoder.com/mathematicians-say-openai-doesnt-understand-its-own-navier-stokes-proof/ - [2026-11-14] CASRAI summary of the Yang–Mills claim: the Gross committee was appointed by OpenAI. It says it does not fully understand the proof and that the mass gap is "untouched". Woit argues that the verified existence is not the standard the Clay problem requires. Open items remain on whether the Lean statement matches Clay's existence requirement and whether the proof can be made human-readable. This is the only post-12-September claim of substance, and it is explicitly half a problem. Partial results are excluded by the rules. https://casrai.org/news/ai-rivals-physics-mathematical-challenge-2026 - [2026-11-12] Dataconomy: the 1,147-page written Yang–Mills proof was still unreleased. The reviewers said the work does not prove uniqueness of the infinite-volume limit. A written work is also needed for the "specific publicly available written work" requirement of the Source Agency route. https://dataconomy.com/2026/11/12/nobel-panel-validates-openais-yang-mills-existence-proof/ - [2026-11-12] Scientific American and Yahoo report that the Yang–Mills mass gap "remains unresolved". OpenAI's own timeline has the written proof from 9 September and QCD from 18 September, but only the 28 October Lean submission was assessed. A full Yang–Mills claim would need a mass-gap proof, written and checked, within about 6 weeks. No such claim exists. https://tech.yahoo.com/science/articles/openai-shepherds-affirm-blockbuster-physics-123000155.html - [2026-11-16] Wikipedia (2026) states that Clay does not consider proposed solutions until at least two years after publication. It also says the OpenAI Navier–Stokes result is unverified and subject to a priority dispute. Nothing in the news shows CMI or a recognized body ruling since 11 September. https://en.wikipedia.org/wiki/Millennium_Prize_Problems - [2026-11-16] Precedent: Perelman's 2002–03 preprints were recognised in 2010. The two-year clock started only when others published expositions. Science News (2010) says the delay exists to let the community find errors. Even a correct AI proof is likely to follow this pattern. https://www.sciencenews.org/article/million-dollar-math-prize-awarded-not-necessarily-accepted - [2026-11-12] Mechanism for the "two Source Agencies" route: for this to fire on a new problem, two outlets would need to report that it is "resolved", citing a written work. Current coverage uses hedged wording ("existence part", "half", "mass gap open"). Outlets also report that OpenAI's other Millennium-related output is partial. Journalists have been careful, so a premature false positive is unlikely. https://www.scientificamerican.com/article/openais-shepherds-affirm-blockbuster-physics-result/ BOTTOM LINE: With 6 weeks left, the only post-12-September claim (Yang–Mills) is admittedly half-proved. The Navier–Stokes claim predates the market and is under a mismatch/error dispute, and Clay's two-year rule and journal timelines rule out the formal routes. The status quo is No unless a new complete AI proof appears and is accepted almost immediately. NOT FOUND: Any Clay or recognized-body statement after 2026-09-11. Any claim of a full mass-gap, P vs NP, Riemann, Hodge or BSD proof. Any journal acceptance of an AI Millennium result. The market's list of Source Agencies.
3. Forecasters 2 draws, Opus 5.5 + Fable 5.2
claude-opus-5-5, high effort
62% conf
Yes 9%
No 91%
already resolved?
Nothing has resolved: CMI lists Poincaré as the only solved problem, Navier–Stokes as 'Active' and five problems as 'Unsolved'. The only post-Sept-12 claim, OpenAI's Yang–Mills work, proves existence only; the Gross committee says the mass gap is untouched, so it is a partial result under the rules. The Riemann, Hodge and BSD releases are explicitly partial. Navier–Stokes was announced Sept 8, before the market opened, is excluded by 'another', and is clouded by the KCL Lean-mismatch preprint and a priority dispute.
time left
About 6 weeks remain; Clay's two-year rule and journal timelines block the formal routes, leaving only the two-Source-Agency route.
if nothing changes
The status quo is No.
path to no
The mass gap stays open, partial results keep coming, and press framing stays hedged through year-end.
path to yes
OpenAI's fast pipeline produces a full mass-gap proof or another complete problem in December, and two major outlets report it resolved citing a public written work; a lenient Navier–Stokes reading is a minor extra path. The Kalshi mid is 15% and Polymarket's 0-more contract implies about 13%; I shade slightly lower because the recent bump reflects committee news that does not meet the rules and experts see no clear path to the mass gap.
claude-fable-5-2, high effort
62% conf
Yes 11%
No 89%
already resolved?
The resolving condition is not met. The CMI page lists only Poincaré as solved, Navier–Stokes as 'Active' and the other five as 'Unsolved', and no journal publication, recognised-body statement or pair of Source Agency reports of a complete new solution was found. Navier–Stokes was claimed on Sept 8, before the Sept 12 issuance, so the 'another' wording most naturally excludes it, and it also faces a Lean-versus-written-proof mismatch dispute.
time left
About 6.5 weeks remain until Jan 1, 2027.
if nothing changes
Status quo is No: the only substantive post-issuance claim, OpenAI's Yang–Mills result, covers existence only. The Gross committee and Scientific American both say the mass gap is untouched, and partial results are excluded by the rules.
path to no
No complete mass-gap, Riemann, P vs NP, Hodge or BSD proof appears by year-end, or one appears but outlets hedge it as a 'claim' or an error is flagged. Clay's two-year rule and journal refereeing times close the formal routes, leaving only the two-Source-Agency route.
path to yes
OpenAI or Anthropic, which have been releasing results in bursts on an unreleased model, announce a full Yang–Mills or other proof in the next few weeks and two Source Agencies report it as solved, as Quanta's headline did for Navier–Stokes. A lenient resolver counting a later Navier–Stokes recognition event is a smaller path. Kalshi is at 15% mid (12/18, last trade 12%) and Polymarket's equivalent implies about 13%. The research shows no pending complete claim and an expert quoted in Science sees no clear path to the mass gap, so I shade slightly below the market while keeping tail weight for the pace of AI results and the rule ambiguity.
4. Final plain mean of 2 draws
10% (62% confidence)
Yes: 10% No: 90%
average of 2 draws (9%, 11%)
already resolved?
Nothing has resolved: CMI lists Poincaré as the only solved problem, Navier–Stokes as 'Active' and five problems as 'Unsolved'. The only post-Sept-12 claim, OpenAI's Yang–Mills work, proves existence only; the Gross committee says the mass gap is untouched, so it is a partial result under the rules. The Riemann, Hodge and BSD releases are explicitly partial. Navier–Stokes was announced Sept 8, before the market opened, is excluded by 'another', and is clouded by the KCL Lean-mismatch preprint and a priority dispute.
time left
About 6 weeks remain; Clay's two-year rule and journal timelines block the formal routes, leaving only the two-Source-Agency route.
if nothing changes
The status quo is No.
path to no
The mass gap stays open, partial results keep coming, and press framing stays hedged through year-end.
path to yes
OpenAI's fast pipeline produces a full mass-gap proof or another complete problem in December, and two major outlets report it resolved citing a public written work; a lenient Navier–Stokes reading is a minor extra path. The Kalshi mid is 15% and Polymarket's 0-more contract implies about 13%; I shade slightly lower because the recent bump reflects committee news that does not meet the rules and experts see no clear path to the mass gap.
Pipeline Timing and Cost
Total pipeline time: 74.8s
LLM cost: $1.40 (Sonnet 5.5 $0.66, Opus 5.5 $0.30, Fable 5.2 $0.23, 21 web searches $0.21)
Per-tool research timings shown in the Research section above.