forecasting
agent
← all forecasts

Will the Riemann Hypothesis be resolved before 2027?

kalshiScience and Technology2026-10-28llm cost $1.01view on kalshi ↗
5%
Agent
10%
Market Price
-5.0pp
Edge
71%
Confidence
2,812
Volume
3.0c
Spread
71
Days to resolution
6
Markets in event
final rationale
average of 2 draws (5%, 5%)
already resolved?
The condition is not met. Clay's official page (fetched, ©2026) still lists the Riemann Hypothesis as "Unsolved". No peer-reviewed proof, authoritative recognition, or Source Agency reporting of a full proof exists. The live developments are partial results that do not count under the original-form clause: OpenAI's quasi-RH strip (1/8–7/8, or 1/12–11/12) and Anthropic's 67.2% critical-line bound, which Anthropic says is not a solution.
time left
Only about 9 weeks remain before the Jan 1, 2027 cutoff.
if nothing changes
If nothing changes, the market resolves No. Journal review (about a year at Annals) and recognition by Clay, AMS or IMU (years) cannot happen in time, so the two-Source-Agency route is the only feasible path to Yes.
path to no
Experts keep digesting quasi-RH and no full-RH claim appears. Or a claim appears, but outlets report it only as "claimed", or an identified error goes unaddressed.
path to yes
An AI lab posts a full critical-line proof in November or December. Rapid human and Lean verification, as with quasi-RH, follows, and Quanta and Science report RH as proven with no flagged errors.
The frontier is moving unusually fast in this timeline, with quasi-RH, no Siegel zeros and a Navier-Stokes claim within weeks, so the tail is real rather than negligible. Even so, experts doubt the quasi-RH technique extends to the 1/2 line. Polymarket prices a CMI declaration by end-2026 at about 2% and "further substantive progress" by end-2026 at about 7.5%. I shade below Kalshi's 10% (bid 8 / ask 11) to about 5%, since the market likely includes a longshot premium.
price vs. prediction
market price agent forecast at t+0
24 follow-ups
earlier forecasts on this market
datepredictedmarket priceconfidence
2026-10-10 6% 8% 71%
1. Planner Opus 5.5
Tools chosen: kalshi_direct resolution_source kalshi_related polymarket_related gdelt_news
Planner reasoning
The Riemann Hypothesis is extremely unlikely to be resolved within about two months, so the Kalshi price is the main anchor. Related markets for other Millennium Prize problems give cross-venue context. The resolution source is Clay's prize page, and Wikipedia covers any recent claimed proofs, so one targeted news check for a serious claim is enough.
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** (KXRIEMANNRES-40-27JAN01) - Current price (probability): 10.00% (live; bid 8% / ask 11%) - No daily close yet, so no price history - resolution_source: already_resolved=unclear; matches_criteria=True; last_updated=unknown (page footer © 2026). The Clay Mathematics Institute's official Riemann Hypothesis page (copyright 2026) still labels the problem "Unsolved". It says the hypothesis has been checked for the first 10,000,000,000,000 zeros. Recent news covers only partial results, not a proof of RH as Clay states it: Quanta (Oct 19, 2026) reports - kalshi_related: 0 related markets / summaries. series KXRIEMANNRES: | keyword 'Millennium Prize': no matches | keyword 'conjecture resolved': no matches - polymarket_related: Polymarket search over 2 keyword(s), kept 10 matches. keyword 'Riemann Hypothesis': 5 markets | keyword 'Millennium Prize': 5 markets - gdelt_news: GDELT: 10 articles across 1 queries (lookback=120d). 'Riemann Hypothesis proof claimed 2026': 10 hits Research loops, bottom lines - current state: As of 2026-10-28, no credible full proof or disproof of RH exists. The only major live developments are AI-generated partial results (the quasi-RH zero-free region and a 67.2% critical-line proportion) that are still being checked. Those results fall short of the actual hypothesis, and the sources I found treat full RH as still open. - base rate: The historical base rate is about 0 resolutions in 167 years, and 1 of 7 Millennium problems resolved in 26 years. Even Perelman's accepted proof took years to be confirmed. The 2026 AI results (quasi-RH, the 67% critical-line bound) are partial, and no full proof is claimed by any credible source. Other markets price a CMI declaration by 2026-12-31 at about 2–3%. The looser Kalshi criteria, which can be met by two media reports citing a public proof, leave a small extra tail. - yes case: The YES case rests on a fast-moving AI frontier with the quasi-RH proof's human verification as a template, plus a loose "two reports" clause. It needs a full-RH claim that does not yet exist, and experts are unsure whether the quasi-RH method extends to it. - no case: Under the literal criteria, a Yes needs a full RH proof (not quasi-RH) that is either published after refereeing, recognized by a body, or corroborated by two outlets with no flagged errors, all within about nine weeks. No credible full-RH claim exists, and every institutional and journal timeline is far longer than that window. Status-quo inertia strongly favours No, and the main residual risk is a surprise AI release that the media would report as a proof.
tools
toolstatustimeitemssummary
kalshi_direct OK 33.3s 1 ## This Market's Kalshi Data **Before 2027** (KXRIEMANNRES-40-27JAN01) - Current price (probability): 10.00% (live; bid 8% / ask 11%) - No daily close yet, so no price history
resolution_source OK 10.6s 1 already_resolved=unclear; matches_criteria=True; last_updated=unknown (page footer © 2026). The Clay Mathematics Institute's official Riemann Hypothesis page (copyright 2026) still labels the problem "Unsolved". It says the hypothesis has been checked for the first 10,000,000,000,000 zeros. Recent n
kalshi_related OK 16.6s 0 0 related markets / summaries. series KXRIEMANNRES: | keyword 'Millennium Prize': no matches | keyword 'conjecture resolved': no matches
polymarket_related OK 0.7s 10 Polymarket search over 2 keyword(s), kept 10 matches. keyword 'Riemann Hypothesis': 5 markets | keyword 'Millennium Prize': 5 markets
gdelt_news OK 28.8s 10 GDELT: 10 articles across 1 queries (lookback=120d). 'Riemann Hypothesis proof claimed 2026': 10 hits
web research agents 4/4 ok, 16 searches, 2 page reads
dimensionstatustimesearchespage readsreport
current state OK 51s 4 0 4354 chars
base rate OK 94s 4 0 5103 chars
yes case OK 25s 4 0 5886 chars
no case OK 24s 4 2 6028 chars
report: current state
- [2026-10-28] The Clay Mathematics Institute's Millennium Problems page lists the Riemann Hypothesis under "Unsolved problems". Navier-Stokes is separately listed as "Active". Only Poincaré is listed as solved. https://www.claymath.org/millennium-problems/ - [2026-10-27] Terence Tao's blog says Maksym Radziwiłł would give a hybrid talk on the quasi-Riemann hypothesis at Courant on 2026-10-28. This is a human-expert review event, not a claim of RH. https://terrytao.wordpress.com/2026/10/27/the-quasi-riemann-hypothesis/ - [2026-10-19] Quanta reports that OpenAI's 2026-10-06 release included a claimed proof of the "quasi-Riemann hypothesis". This is a weaker statement: zeros lie in a strip narrower than (0,1), not on the 1/2 line. https://www.quantamagazine.org/ai-proved-the-quasi-riemann-hypothesis-what-happens-now-20261019/ - Version 1 is dated 2026-09-30. It claims zeros lie between 1/8 and 7/8. It is 199 pages and Lean-verified. - Version 2 claims 1/12 to 11/12. It is 49 pages and not yet formally verified. - Radziwiłł said both papers are badly written and would be rejected immediately by a journal. - Number theorists are skeptical that the technique extends to full RH. The Quanta text was cut off after that point. - [2026-10-14 approx.] Science reports that OpenAI's latest proof solves only a weaker version of RH. It reports Maynard calling the result "groundbreaking". https://www.science.org/content/article/ai-edges-closer-solving-riemann-hypothesis-one-math-s-hardest-problems - [2026-10-06] OpenAI released 722 manuscripts on GitHub covering 377 problems. They included partial results on RH, Birch and Swinnerton-Dyer, and Hodge. Mathematicians raised concerns about validity and exposition. Peter Sarnak said he would not call the proof a "theorem" until humans verify it. https://en.wikipedia.org/wiki/OpenAI_bulk_release_of_mathematical_papers - [2026-10 (undated)] Scientific American says it is "too early to say" whether the quasi-RH proof leads toward full RH. Hector Pasten is quoted saying that some people claim it is progress toward Riemann and others say it is not. https://www.scientificamerican.com/article/the-most-exciting-claims-from-openais-heap-of-new-proofs/ - [2026-08-10] Anthropic reported that an unreleased Claude model pushed a bound on the proportion of zeros that are simple and on the critical line from 41.6% to 67.2%. The company said it did not solve RH and that the method is unlikely to prove the full hypothesis. As of 2026-08-10 the result had not been through journal peer review. https://kingy.ai/blog/claude-riemann-hypothesis-67-percent-result/ - [2026-10-27] 3 Quarks Daily re-posted the Quanta article. It describes the quasi-RH result as a weaker version of RH and says RH has been open for over 160 years. https://3quarksdaily.com/3quarksdaily/2026/10/ai-proved-the-quasi-riemann-hypothesis-what-happens-now.html - [2026-01-01 and later] Many fringe "proofs" of RH are circulating, for example a Robert Edward Grant paper dated 2026-01-01 and preprints on Cambridge Open Engage. None is peer-reviewed or recognized. https://robertedwardgrant.com/proof-of-the-riemann-hypothesis/ - Rule context: Clay's rules, as summarized in one secondary source, require publication in a peer-reviewed journal followed by a two-year waiting period. Clay's own prize rules are not met by any item found. This differs from the Kalshi criteria, which can resolve on a published peer-reviewed article or on two Source Agency reports. https://baike.baidu.com/en/item/Millennium%20Prize%20Problems/1526093 BOTTOM LINE: As of 2026-10-28, no credible full proof or disproof of RH exists. The only major live developments are AI-generated partial results (the quasi-RH zero-free region and a 67.2% critical-line proportion) that are still being checked. Those results fall short of the actual hypothesis, and the sources I found treat full RH as still open. NOT FOUND: Any Clay, AMS, IMU or journal statement recognizing a resolution of RH. Any credible full-RH proof claim endorsed by experts. A scheduled event before 2027-01-07 that could resolve RH. Independent journal acceptance of the OpenAI or Anthropic results. I did not fully read the Science, Scientific American, Wikipedia or kingy.ai pages. Those bullets rest on search snippets, and the Quanta page was truncated in the middle.
report: base rate
- [2026-10-28] Base rate: RH has been open for 167 years since 1859. Of the 7 Millennium Problems, only the Poincaré conjecture is officially declared solved. Perelman posted it in 2002–03, and Clay announced the prize on 2010-03-18. That is about 7 years from posting to Clay's declaration, and no journal publication was needed. https://en.wikipedia.org/wiki/Millennium_Prize_Problems - [2026-10-19] Quanta: OpenAI announced 377 results on 2026-10-06, including a "quasi-Riemann hypothesis". This is a weaker claim, with zeros confined to a strip narrower than (0,1). The first version (dated 2026-09-30, 199 pp) puts zeros in 1/8–7/8 and is Lean-verified. The second version (49 pp) gives 1/12–11/12 and is not yet formally verified. Experts call it a major result, but it is not full RH. https://www.quantamagazine.org/ai-proved-the-quasi-riemann-hypothesis-what-happens-now-20261019/ - [2026-10-09] Scientific American, via a snippet I did not fetch: Pasten says it is too early to say whether quasi-RH is progress toward full RH. "Some people claim it is progress toward Riemann; other people say it's not." https://www.scientificamerican.com/article/the-most-exciting-claims-from-openais-heap-of-new-proofs/ - [2026-08] Anthropic, per a secondary source (tech-insider, snippet only): an unreleased Claude model raised a proven lower bound for the fraction of zeros on the critical line (41.6% to 67.2%), formalized in Lean. Anthropic explicitly did not claim to have solved RH. https://tech-insider.org/anthropic-claude-riemann-hypothesis-breakthrough-2026/ - [2026-10-28] Polymarket "Will CMI declare RH solved by…" (opened 2026-08-13), page rendered on 2026-10-28. By 2026-12-31 the price is about 2–3% (Yes 2.9¢). By 2027-12-31 it is 11%. Volume is about $40K. This is a stricter test than the Kalshi contract, because it requires a CMI declaration and a public proof alone does not count. Its context text says historical precedent requires years of scrutiny before any declaration. https://polymarket.com/event/will-cmi-declare-the-riemann-hypothesis-solved-by-20260812114814549 - [2026-10-20] Polymarket "Further substantive progress on RH by…" is about 12–13% for 2026-12-31 and about 3–5% for 2026-10-31. Its context says the OpenAI quasi-RH result does not meet the market's threshold, and that full resolution "remains distant". https://polymarket.com/event/further-substantive-progress-on-the-riemann-hypothesis-by-20260812121554604 - [2026-09] Octagon AI aggregator (snippet): the market-implied chance of RH being solved before 2028 was about 20%, and its own model said about 12%. Before 2030 was 44% market versus 28% model. Before 2035 was 56% market versus 38% model. The underlying market is unnamed, probably Polymarket or Manifold, so treat this as low reliability. https://www.octagonai.co/markets/science-and-technology/physics-math/when-will-the-riemann-hypothesis-be-solved - [2026-08-06] Manifold "AI program resolves RH before 2035" was at 44%. This needs an AI-authored proof vetted to Annals standard. Manifold is play money. https://tech-insider.org/ai-riemann-hypothesis-2035-manifold-market-odds/ - [2026-10] Manifold sibling markets (snippet, undated): "OpenAI discovers the first proof of RH before 2027?" is 14%. "Will RH be solved before 2030?" is 80%. These two look inconsistent, probably because of differing resolution terms, so I treat them as unreliable. A Manifold comment argues a proof would need about 6 months of peer review plus 6 months of expert review before acceptance. https://manifold.markets/HarrisonLucas/ai-program-resolves-riemann-hypothe - [2026-01-01 onward] Crank and claimed proofs appear continually, for example Grant (2026-01-01) and Abe's Cambridge Open Engage preprint (v19, 2025-11-27). None has been accepted. The Kalshi criteria require peer-reviewed publication, or an authoritative body's recognition, or two Source Agencies reporting a proof with no unaddressed identified errors. https://robertedwardgrant.com/proof-of-the-riemann-hypothesis/ - [2026-10-28] I found no report that RH has been proved or disproved by a credible source. The 2026-10-06 OpenAI Navier–Stokes counterexample claim is a separate problem. Wikipedia still lists only Poincaré as solved. https://en.wikipedia.org/wiki/Millennium_Prize_Problems BOTTOM LINE: The historical base rate is about 0 resolutions in 167 years, and 1 of 7 Millennium problems resolved in 26 years. Even Perelman's accepted proof took years to be confirmed. The 2026 AI results (quasi-RH, the 67% critical-line bound) are partial, and no full proof is claimed by any credible source. Other markets price a CMI declaration by 2026-12-31 at about 2–3%. The looser Kalshi criteria, which can be met by two media reports citing a public proof, leave a small extra tail. NOT FOUND: Any Metaculus RH question page or a numeric Metaculus forecast. Any formal expert poll with a track record. Any CMI statement on the OpenAI results. The Kalshi contract's own price, which I did not search for, as instructed. I only read the second half of the SciAm and Octagon pages via snippets.
report: yes case
YES means: the Riemann Hypothesis, in its Clay formulation, is resolved before 2027. That requires a peer-reviewed proof or disproof, a recognition by an authoritative body, or two Source Agencies reporting a proof with no unaddressed error. Resolution must stand unretracted at expiry. - [2026-10-19] A Quanta report shows human experts are engaging with the AI quasi-RH proof, not dismissing it. Sarnak repurposed a course to study it. Radziwiłł can sketch its argument in half a page, and other write-ups run 7–13 pages. A secondary summary says experts increasingly believe the proof is correct. This matters because the fast-verification pipeline (humans plus Lean) now exists for a closely related object. https://www.quantamagazine.org/ai-proved-the-quasi-riemann-hypothesis-what-happens-now-20261019/ - [2026-10-19] Quanta quotes Radziwiłł saying the vast majority of consequences of RH already follow from quasi-RH. The proof needed no new concepts. It linked old ones in unexpected ways. This is the strongest YES mechanism: if the method is a synthesis of known tools, a similar synthesis could in principle be pushed to the 1/2 line. No source says it can. https://www.quantamagazine.org/ai-proved-the-quasi-riemann-hypothesis-what-happens-now-20261019/ - [2026-10-06] Sources describe a steep release cadence for the OpenAI model. About 10 results in August, over 100 in September, and 722 by Oct 6. A further drop could include a full-RH claim before year-end. In that case the third criterion could be met quickly, because it needs only two Source Agencies reporting a "proof" tied to a public work. Reporting speed is a tail risk, not a sign of correctness. https://www.mindstudio.ai/blog/openai-722-math-proofs-release - [2026-10-06] Sloppy headlines already exist. A KuCoin news item claimed OpenAI's model "has proven the Riemann Hypothesis". It also described the Lean verification as covering that claim. Other outlets say the claim is only quasi-RH. Careless outlets could still trip the "two Source Agencies" clause, but the clause also requires the reporting to describe the original-form conjecture as resolved. https://www.kucoin.com/news/flash/openai-solves-722-math-problems-quasi-riemann-hypothesis-proven - [2026-10-13] Science says Maynard, who made the most recent notable Riemann advance in 2024, calls the result "groundbreaking". Science also notes the release came less than a month after OpenAI's Navier–Stokes announcement. So there are two Millennium-class claims inside about a month, which favors a fast-moving frontier. https://www.science.org/content/article/ai-edges-closer-solving-riemann-hypothesis-one-math-s-hardest-problems - [2026-10-09] The same release has a second result on Siegel zeros. It shows that zeta and many Dirichlet L-functions have no Siegel zero, which Pasten says "settles the question". The release therefore contains two Riemann-adjacent results, not one. https://www.scientificamerican.com/article/the-most-exciting-claims-from-openais-heap-of-new-proofs/ - [2026-10-27] Tao's blog top-posts list includes "A sketch of the proof of non-existence of Siegel zeroes" and "Simple and critical zeros of the Riemann zeta-function". Radziwiłł's Courant talk is on 2026-10-28. Leading human mathematicians are digesting and rewriting the AI work within weeks. That supports fast verification of AI-generated number theory. https://terrytao.wordpress.com/2026/10/27/the-quasi-riemann-hypothesis/ - [2026-09-09] Octagon's trace of a Kalshi-linked RH market shows it moved from about 10% (Aug 27) to a peak of 18% (Sep 5) on the Claude 67.2% result. It then slipped to 17% by Sep 9. Traders have treated AI news as real evidence. These are longer-dated strikes (before 2028 and later), not this contract. https://www.octagonai.co/markets/science-and-technology/physics-math/when-will-the-riemann-hypothesis-be-solved - [2026-10-28] Kalshi's sibling "Millennium Prize" ladder (page about 4 days old) shows Before 2028 at 8%, Before 2030 at 32%, Before 2032 at 50%. The "Before 2028" strike shows a 5% chance on another snapshot. Even the prize-award market, which needs Clay's slow process, prices a material but small near-term probability. https://kalshi.com/markets/kxriemann/riemann-hypothesis-solved/kxriemann-35 - [2026-10-06] Wikipedia records Sarnak declining to call the quasi-RH proof a "theorem" until humans verify and understand it. Maynard said the AI put together non-mainstream ideas "in an exceptionally clever way". The quasi-RH result is the only real near-RH object, and its verification is the only credible leading indicator. https://en.wikipedia.org/wiki/OpenAI_bulk_release_of_mathematical_papers **Single most likely path to YES:** 1. OpenAI, Anthropic or a similar lab posts a full 1/2-line RH proof (or a counterexample) on GitHub or arXiv before about December 2026. 2. Humans and Lean verify it within days. The quasi-RH proof drew expert unpacking in about two weeks. 3. Two Source Agencies (for example Quanta and Science) report it as proven, citing the public manuscript. 4. No identified error goes unaddressed before expiry. I found no evidence that step 1 has occurred. Labs describe their own results as partial. Anthropic's August work explicitly disclaimed solving RH. BOTTOM LINE: The YES case rests on a fast-moving AI frontier with the quasi-RH proof's human verification as a template, plus a loose "two reports" clause. It needs a full-RH claim that does not yet exist, and experts are unsure whether the quasi-RH method extends to it. NOT FOUND: Any credible full-RH proof or disproof claim, in any form. Any lab statement about an imminent full-RH release. Any Clay, AMS, IMU or journal recognition. Any journal submission or acceptance of the quasi-RH papers. The truncated middle of the Quanta article, which covers expert views on extension to full RH.
report: no case
YES means: the Riemann Hypothesis in its original Clay form is resolved before 2027-01-01. That needs a peer-reviewed published proof or disproof, or a written recognition by an authoritative body, or two Source Agencies reporting a proof with no unaddressed error. - [2026-10-28] Clay's own rules (fetched) bar CMI from considering any solution until it is published in a Qualifying Outlet, two years have passed, and it has general acceptance. CMI does not take direct submissions. Even a proof published tomorrow would not reach CMI's consideration stage until late 2028. This does not bind the Kalshi criteria, but it shows how slowly the community validates things. https://www.claymath.org/millennium-problems/rules/ - [2026-09-11] CMI described the Navier-Stokes claim as "apparently" settled but awarded nothing and kept it "Active". This is the closest recent precedent for an AI claim on a Millennium problem. Even a loud, verified-in-Lean AI claim produced only a hedge from CMI within about two weeks. https://mathworld.wolfram.com/MillenniumPrizeProblems.html - [2026-10-28] The only live RH-related claim is the quasi-RH proof, and it is not the RH. Its two forms put zeros in strips of 1/8–7/8 and 1/12–11/12. Neither gives Re(s)=1/2, so the Kalshi "original form" clause is not met. Quasi-RH being accepted would not make a "complete proof" of RH. Scientific American says it is too early to tell whether quasi-RH leads to the full hypothesis. https://www.scientificamerican.com/article/the-most-exciting-claims-from-openais-heap-of-new-proofs/ - [2026-10-19] The fetched Quanta article adds that the human write-ups of the quasi-RH proof (7–13 pages) are "circulating privately" and unpublished. It says Radziwiłł called the papers unfit for any journal. The 11/12 paper is the only one of the 722 that a human edited, and OpenAI will not name the editor. So even the partial result has no journal submission, so far as I could find. Sarnak is spending a semester just on understanding it. https://www.quantamagazine.org/ai-proved-the-quasi-riemann-hypothesis-what-happens-now-20261019/ - [2026-10-28] Journal timelines make a Yes through publication nearly impossible. One tracker (Apr 2026) estimates about 42 weeks median review at Annals of Mathematics, plus about 14 weeks of production. That is about 56 weeks in total, against about 9 weeks to the contract's 2027-01-01 cutoff. https://www.getpaperpin.com/journals/annals-of-mathematics - [2026-10-28] The "authoritative body" route needs a written communication that cites a specific written work. AMS, IMU and the academies move on timescales of years. CMI itself took until 2010 for Poincaré. The 2018 rules also require two years. I found no precedent for such a body recognizing a Millennium result within months. https://www.sciencenews.org/article/million-dollar-math-prize-awarded-not-necessarily-accepted - [2026-10-28] The media route is the only fast path, and it carries a built-in veto. It requires two Source Agencies describing RH as proven or resolved. It also requires no reported unaddressed identified error. Major outlets (Quanta, Science, SciAm) are carefully labelling the AI results as a weaker version, and Science's headline says "edges closer". Fringe claims (Quinones, Cambridge Open Engage) are not covered by credible agencies. https://www.science.org/content/article/ai-edges-closer-solving-riemann-hypothesis-one-math-s-hardest-problems - [2026-10-28] Past high-profile RH claims have all died. Atiyah (2018) sent his proof to 15 experts and got skepticism or silence, and it was never printed. In the Easwaran case, CMI president Bridson publicly told media not to call RH proven. The Atiyah episode shows that, in the "two Source Agencies" route, reporting that a claim was merely "claimed" does not satisfy the clause, and expert pushback follows quickly. https://www.science.org/content/article/skepticism-surrounds-renowned-mathematician-s-attempted-proof-160-year-old-hypothesis - [2026-09 approx.] Experts who studied the Anthropic 67.2% critical-line result describe it as a Lean-formalized but unreviewed partial advance. One explainer says it does not move RH "67% of the way" and says nothing about the remaining third of zeros. This is a lower-bound improvement, not a route to completion. https://www.mathlumen.com/articles/riemann-hypothesis-what-is-actually-proven-and-why-its-hard - [2026-10-27] Status-quo evidence: on the eve of the Courant talk, Tao's blog post concerns quasi-RH only. It makes no suggestion that full RH is near. https://terrytao.wordpress.com/2026/10/27/the-quasi-riemann-hypothesis/ - [2026-10-28] The AI-capability tail runs in only one direction. OpenAI went from a Navier-Stokes claim (Sept) to 377 results (Oct 6) in about a month, so a surprise "full RH" drop is conceivable. But any such drop would need to be both claimed and corroborated within about 9 weeks. Skeptical experts say the quasi-RH technique does not obviously extend (SciAm quotes Pasten saying people disagree). If OpenAI did claim full RH, the "unaddressed identified error" clause would likely be triggered by the first expert critique. https://en.wikipedia.org/wiki/Millennium_Prize_Problems BOTTOM LINE: Under the literal criteria, a Yes needs a full RH proof (not quasi-RH) that is either published after refereeing, recognized by a body, or corroborated by two outlets with no flagged errors, all within about nine weeks. No credible full-RH claim exists, and every institutional and journal timeline is far longer than that window. Status-quo inertia strongly favours No, and the main residual risk is a surprise AI release that the media would report as a proof. NOT FOUND: Any journal submission or acceptance of the OpenAI quasi-RH or Anthropic results. Any statement from AMS, IMU or CMI on RH or the OpenAI results. Any credible full-RH proof claim. The full text of the CMI rules PDF (the web page summary only). The remainder of the Quanta article after Sarnak's quote (it was truncated again).
3. Forecasters 2 draws, Opus 5.5
claude-opus-5-5, high effort
72% conf
Yes 5%
No 95%
already resolved?
The condition is not met. Clay's official page (fetched, ©2026) still lists the Riemann Hypothesis as "Unsolved". No peer-reviewed proof, authoritative recognition, or Source Agency reporting of a full proof exists. The live developments are partial results that do not count under the original-form clause: OpenAI's quasi-RH strip (1/8–7/8, or 1/12–11/12) and Anthropic's 67.2% critical-line bound, which Anthropic says is not a solution.
time left
Only about 9 weeks remain before the Jan 1, 2027 cutoff.
if nothing changes
If nothing changes, the market resolves No. Journal review (about a year at Annals) and recognition by Clay, AMS or IMU (years) cannot happen in time, so the two-Source-Agency route is the only feasible path to Yes.
path to no
Experts keep digesting quasi-RH and no full-RH claim appears. Or a claim appears, but outlets report it only as "claimed", or an identified error goes unaddressed.
path to yes
An AI lab posts a full critical-line proof in November or December. Rapid human and Lean verification, as with quasi-RH, follows, and Quanta and Science report RH as proven with no flagged errors.
The frontier is moving unusually fast in this timeline, with quasi-RH, no Siegel zeros and a Navier-Stokes claim within weeks, so the tail is real rather than negligible. Even so, experts doubt the quasi-RH technique extends to the 1/2 line. Polymarket prices a CMI declaration by end-2026 at about 2% and "further substantive progress" by end-2026 at about 7.5%. I shade below Kalshi's 10% (bid 8 / ask 11) to about 5%, since the market likely includes a longshot premium.
claude-opus-5-5, high effort
70% conf
Yes 5%
No 95%
already resolved?
Not met: Clay's official page still lists RH as 'Unsolved'. No peer-reviewed proof, no authoritative-body recognition and no Source Agency reporting of a full-RH proof was found.
time left
About 9 weeks remain until the 2027-01-01 cutoff, so peer-review and institutional routes are effectively closed and only the 'two Source Agencies' media route is plausible.
if nothing changes
Status quo is No.
path to no
The AI results stay partial (OpenAI's quasi-RH strip of 1/8–7/8 or 1/12–11/12, Anthropic's 67.2% critical-line bound), experts keep digesting them, outlets keep labelling them weaker versions, and no full-RH claim appears by year-end.
path to yes
OpenAI or another lab drops a Lean-verified full proof, Quanta and Science report it as proven, and no unaddressed error surfaces before expiry. The Kalshi anchor of 10% looks rich given no full-RH claim exists, experts doubt the quasi-RH method extends, the error-veto clause would likely bite, and related Polymarket markets price a CMI declaration by end-2026 at about 2% and 'substantive progress' at about 7.5%. The remarkably fast AI frontier justifies a non-trivial tail, so 5%.
4. Final plain mean of 2 draws
5% (71% confidence)
Yes: 5% No: 95%
average of 2 draws (5%, 5%)
already resolved?
The condition is not met. Clay's official page (fetched, ©2026) still lists the Riemann Hypothesis as "Unsolved". No peer-reviewed proof, authoritative recognition, or Source Agency reporting of a full proof exists. The live developments are partial results that do not count under the original-form clause: OpenAI's quasi-RH strip (1/8–7/8, or 1/12–11/12) and Anthropic's 67.2% critical-line bound, which Anthropic says is not a solution.
time left
Only about 9 weeks remain before the Jan 1, 2027 cutoff.
if nothing changes
If nothing changes, the market resolves No. Journal review (about a year at Annals) and recognition by Clay, AMS or IMU (years) cannot happen in time, so the two-Source-Agency route is the only feasible path to Yes.
path to no
Experts keep digesting quasi-RH and no full-RH claim appears. Or a claim appears, but outlets report it only as "claimed", or an identified error goes unaddressed.
path to yes
An AI lab posts a full critical-line proof in November or December. Rapid human and Lean verification, as with quasi-RH, follows, and Quanta and Science report RH as proven with no flagged errors.
The frontier is moving unusually fast in this timeline, with quasi-RH, no Siegel zeros and a Navier-Stokes claim within weeks, so the tail is real rather than negligible. Even so, experts doubt the quasi-RH technique extends to the 1/2 line. Polymarket prices a CMI declaration by end-2026 at about 2% and "further substantive progress" by end-2026 at about 7.5%. I shade below Kalshi's 10% (bid 8 / ask 11) to about 5%, since the market likely includes a longshot premium.
5. Entry Decision Opus 5.5 trader
Side
nobetting no, wins if this resolves no
Position Size
$500
Contracts
543
Trader Confidence
71%
Chance of Yes
  • market at entry 10%
  • agent at entry 5%
  • our side
Agent put yes at 5% with the market at 10%, so we bought no at $0.92 (5.0pp edge).
Trade rationale
This bet is with the favourite: both the market (10%) and the forecast (5%) say the Riemann Hypothesis won't be resolved before 2027, so buying NO matches the book's profitable trade type. There is a concrete reason behind it, not just caution. Clay's official page still lists RH as unsolved. Journal review takes about a year and Clay/AMS/IMU recognition takes years, so the only way to YES in about 9 weeks is a new full critical-line proof that two qualifying news outlets report as proven with no flagged errors. The recent results don't count under the market's original-form clause: OpenAI's quasi-RH strip proves no zeros in a band of the critical strip, not that all zeros lie on the 1/2 line, and Anthropic itself says its 67.2% critical-line bound is not a solution. On Polymarket, a Clay declaration by end-2026 trades at about 2%. Both forecaster draws gave 5% with confidence around 0.71. That's a nudge only, not a gate.
Allocation logic
I'm sizing at the floor because the real edge is thin. Filling NO at about $0.92 against the $0.08 bid shrinks the 5pp edge at the mid to about 3pp, I couldn't check depth (Kalshi's order book wasn't loading), and the price has moved from 8% to 10% against us since the 10/10 scan. That move reflects real AI-math news, and the media-recognition route makes the YES tail fatter than usual.
no contract: entry $0.92, settled at $0.97
Status: WON
P&L: $23.25
Pipeline Timing and Cost
Total pipeline time: 136.0s
LLM cost: $1.01 (Sonnet 5.5 $0.61, Opus 5.5 $0.24, 16 research-loop web searches $0.16)
Per-tool research timings shown in the Research section above.