AI Solving the Riemann Hypothesis: The Manifold Market Everyone Is Betting On Hits 44%. A prediction market asking whether an AI program will resolve the Riemann Hypothesis before 2035 has climbed to a 44% implied probability of YES on Manifold as of August 06, 2026, up from roughly 42% in late July. No proof exists, and the move looks sentiment-driven rather than evidence-driven, powered by a run of AI-assisted math results, including Anthropic’s disclosure that an unreleased Claude model pushed a Riemann-related lower bound from 41.6% to 67.2% in August 2026 — a jump Forbes reported broke a bound that had stood unchanged for 37 years — and a heated debate among mathematicians over what machines should be allowed to claim.
Don't miss new tech stories on Google
Add Tech Insider once in the Google app and our stories appear in your news suggestions.
The Offbeat Market Traders Cannot Stop Watching
The contract in question is titled “AI program resolves Riemann Hypothesis before 2035”, hosted on Manifold, the play-money forecasting platform. It has become one of the most active math-themed markets on the site, drawing roughly 90 unique traders and a steady flow of daily activity. The question is deceptively simple: will software, on its own, crack the single most famous unsolved problem in mathematics within the next nine years?
What makes the market unusual is the gap between the seriousness of the subject and the speculative nature of the wager. There is no confirmed breakthrough. Instead, traders are pricing a probability against a backdrop of rapid, verifiable progress in AI-assisted mathematics, and every new headline nudges the line. For readers new to these venues, our explainer on how prediction markets work covers the basic mechanics.
What the Odds Actually Mean
As of August 06, 2026, the market sits at a 44% implied probability of YES on Manifold. That figure is a crowd estimate, not a measurement. It says the aggregate of participants believes there is a little better than a coin-flip chance against, over a nine-year horizon, that a machine authors an accepted proof. Manifold runs on play money, so the price reflects reputational stakes and forecasting skill rather than large capital at risk.
An implied probability is simply the market price read as a percentage chance. If you are unfamiliar with how these numbers translate into forecasts, see our primer on implied probability. The 44% level is notable mainly because it has drifted upward from about 42%, a roughly two-point climb tracked between July and August 2026 in our earlier report on the same Manifold market. Small on paper, that shift is what put the contract back on trader radar.
The Fine Print in the Resolution Criteria
The market’s terms are strict, which is central to reading the odds correctly. To resolve YES, an AI program must write a correct proof that the Riemann Hypothesis is either true, false, or undecidable within the axioms of complex analysis, before 2035. The proof must be vetted by qualified humans to a standard acceptable to editors of leading journals such as the Annals of Mathematics. Crucially, the resolution language states that the proof “must primarily be the work of the AI” and “cannot be a human-AI collaboration.”
That last clause is the sticking point. Almost every headline AI-math result of the past year has been a collaboration, with a human posing the problem, steering the model, and formalizing the output. A market at 44% is therefore not betting on an imminent announcement. It is betting on nine years of progress reaching a point where a model does the key creative work essentially alone.
What Is Driving the Move
There is no verified claim that any AI has proved the Riemann Hypothesis. Social posts circulating in mid-2026 about a “quantum-AI” system producing a 500-page proof, or a “mathematical singularity,” are not backed by any published, peer-reviewed work, and should be treated as noise. What is documented is narrower but real: Anthropic’s own AI Native report says an unreleased Claude model raised a Riemann Hypothesis-related lower bound from 41.6% to 67.2% in August 2026, a jump Forbes reported the model reached in just 1.5 days while producing what Anthropic described as a formally verifiable result. The most-supported explanation for the drift is broader momentum in AI-assisted mathematics plus an intense professional debate that has kept the topic in the news.
Two threads matter. First, a string of concrete, verified results has shown models contributing to open problems. Second, the mathematics community has responded with formal statements about attribution and integrity, which itself generated coverage. Neither thread proves anything about the Riemann Hypothesis specifically, but together they raise the ambient probability traders assign to fast progress. Attention, in a play-money market, is often enough to move the line.
The Real Results Behind the Hype
The verifiable milestones are genuine and cited by named mathematicians. In January 2026, GPT-5.2 Pro was used to solve several open Erdős problems. One widely covered case, Erdős Problem #281, open since 1980, was tackled by Neel Somani, a former Citadel quantitative researcher, using GPT-5.2 Pro, with the argument then formalized and independently checked. Fields Medalist Terence Tao called it, per The Decoder, “perhaps the most unambiguous instance” of AI solving an open problem to date. Tao also cautioned that these are the “lowest hanging fruit,” problems solvable with standard techniques, and that the win says more about speed than difficulty. Anthropic has separately disclosed, per reporting from Spidits, that an unreleased Claude model tested roughly 650 candidate ideas across 60 parallel subagents by August 2026 while chasing the bound-tightening result described above, an approach that suggests compute-heavy parallel search, not a single flash of insight, is behind the latest gains, a push Forbes noted broke a bound that had stood unchanged for 37 years.
Separately, Google DeepMind reported that a specialized system built on Gemini 2.5 Deep Think, alongside an unpublished math-focused system, helped prove a novel theorem in algebraic geometry with mathematicians including Stanford’s Ravi Vakil. As officechai reported, Vakil described the model’s argument as “the kind of insight I would have been proud to produce myself.” These results explain why the topic feels live. They also underline the distance to the market’s bar, since each involved substantial human direction.
The People and Institutions in the Story
The narrative runs through recognizable figures. Terence Tao has become, in the words of Quanta Magazine on June 08, 2026, an “evangelist for AI in math,” documenting model contributions to the Erdős problem list while stressing a “symbiotic” rather than machine-takeover future. Timothy Gowers, another Fields Medalist, has publicly questioned whether current models can autonomously do research-level mathematics such as deriving hypotheses or building deep proofs.
Institutions are weighing in too. In June 2026, an international group of mathematicians published the Leiden Declaration on Artificial Intelligence and Mathematics, endorsed by the International Mathematical Union and signed by more than 3,000 people. It warns that AI threatens the integrity of proof, the system of attribution, and the autonomy of mathematical research. It was presented at the International Congress of Mathematicians in July 2026. Funding is moving just as fast as the rhetoric: UCLA Samueli announced a $5 million, three-year DARPA contract for AI-for-math research in April 2026, UC Irvine and USC followed with a $2.6 million DARPA AI-math grant in May 2026, and the AI for Math Fund, backed by Renaissance Philanthropy and XTX Markets, added $13.5 million to the field in March 2026. In a July 26, 2026 post on his weblog, Gowers explained that he did not sign, not because he disagreed in any concrete way, but because parts of it make confident assertions he felt uncertain about. That debate kept the AI-and-math story in headlines through the window when the market moved.
The Broader Context for a Tech and Finance Reader
The Riemann Hypothesis, posed in 1859, concerns the distribution of prime numbers and sits at the center of number theory. It is one of the Clay Mathematics Institute’s Millennium Prize Problems, carrying a $1 million reward for a correct proof, one that remains unclaimed as of August 2026, according to The AI Insider. Human progress continues in parallel: in 2024, MIT’s Larry Guth and Oxford’s James Maynard posted an improvement to a long-standing bound, covered by Scientific American as a step closer, though the problem remains open.
For a finance-minded reader, the market is a case study in how forecasting venues price low-frequency, high-uncertainty events. The line reflects a distribution of beliefs about a nine-year trajectory, and it reacts to narrative as much as to fundamentals. That is a feature and a risk. It is also why the same AI-progress story that lifts a math market echoes across model-capability debates covered in our comparison of Claude, Gemini, and ChatGPT. For how these venues differ from regulated exchanges, see our Kalshi versus Polymarket breakdown.
Timeline: AI and Hard Math, 2024 to 2026
| Date | Development | Why it matters |
|---|---|---|
| July 2024 | Guth and Maynard post an improved bound related to the hypothesis | Human progress on a decades-old barrier; problem still open |
| Jan 2026 | GPT-5.2 Pro used to solve open Erdős problems, verified by Tao | First widely accepted cases of AI helping close open problems |
| Jan 2026 | Gemini-based system helps prove a novel algebraic geometry theorem | Model contributes genuine synthesis, praised by named mathematicians |
| June 2026 | Leiden Declaration published, endorsed by the IMU | Formal community stance on AI, proof integrity, and attribution |
| July 2026 | Declaration presented at the ICM; Gowers publishes his response | Debate keeps AI-and-math in the news as the market drifts up |
| Aug 06, 2026 | Manifold market at 44% YES | Crowd prices better-than-even odds against by 2035 |
What to Watch Next
Three signals would matter. First, any peer-reviewed, journal-grade result where a model does the central creative work with minimal human steering, which is what the market’s terms demand. Second, movement on analytic number theory specifically, since Erdős and algebraic geometry wins do not directly bear on the Riemann Hypothesis. Third, how the attribution debate resolves, because if journals adopt rules limiting AI authorship, the “primarily the work of the AI” clause could become harder to satisfy even if capability arrives. Absent a verified proof, expect the line to keep reacting to narrative, drifting on benchmark news and cooling when the hype cycle does.
Frequently Asked Questions
Has an AI actually proved the Riemann Hypothesis? No. There is no published, peer-reviewed proof from any AI or human as of August 06, 2026. Claims circulating on social media about a completed AI proof are unverified.
What does the 44% figure mean? It is the implied probability of YES on the Manifold market as of August 06, 2026. It reflects the crowd’s aggregate estimate that a qualifying AI-authored proof appears before 2035, not any confirmed event.
Why does the market require the proof to be “primarily the work of the AI”? The resolution rules exclude human-AI collaborations. Because nearly all recent AI-math wins involved heavy human direction, this clause sets a high bar that keeps the odds well below certainty.
Is this real money? No. Manifold uses play money, so the price reflects forecasting reputation rather than capital at risk.
What would change the odds fastest? A verified, journal-grade result in analytic number theory where a model does the key creative work with little human input would move the line most.
The Bottom Line
Sources
- Manifold — AI program resolves Riemann Hypothesis before 2035
- The Decoder — Terence Tao says GPT-5.2 Pro cracked an Erdős problem
- officechai — Google Gemini proves novel theorem in algebraic geometry
- Quanta Magazine — How Terry Tao Became an Evangelist for AI in Math
- Leiden University — Leiden Declaration warns AI is challenging the core values of mathematics
- Gowers’s Weblog — Thoughts about the Leiden Declaration
- Scientific American — The Riemann Hypothesis Is a Step Closer to Being Solved
- Clay Mathematics Institute — Riemann Hypothesis Millennium Prize
Prediction markets carry risk and are not investment or betting advice. Market availability is restricted by jurisdiction: Polymarket is not available to US persons, and Kalshi is a CFTC-regulated US exchange. Manifold operates with play money. Participation age limits of 18+ or 21+ apply where relevant. If gambling is a problem for you or someone you know, call 1-800-GAMBLER for confidential help.
