The offbeat Manifold market betting 96% that “something will happen in 2026” is suddenly hot. A deliberately absurd prediction market that asks whether any noteworthy event will occur this year is trading at a 96% implied chance of YES as of August 01, 2026, on Manifold. The reason for the fresh wave of trading is anything but trivial: an artificial-intelligence system helped disprove an 87-year-old mathematics problem in late July, and traders are now arguing over whether that counts.
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 market that asks if anything will happen at all
The market, titled “Will something happen in 2026?” and created by a Manifold user going by Bandors, is a joke question with a serious rulebook. Rather than resolving on a vague vibe, its description ties YES to a long list of concrete triggers. Those include political leadership changes involving figures such as Donald Trump, Vladimir Putin, or Xi Jinping; a Taiwan invasion; a credible claim of artificial general intelligence; a major pandemic; a stock-market crash exceeding 40% from the 2025 peak; the death of a named public figure; and, notably, an AI solving a famous unsolved mathematics problem. That crash trigger looks like the least likely path to YES for now: Goldman Sachs Research’s August 2026 outlook projects 2.8% global GDP growth for the year, ahead of the 2.5% consensus estimate, which is not the kind of backdrop a 40% drawdown typically arrives in.
The premise reads like satire, because almost any year clears at least one of those bars, and 2026 already has non-math candidates lined up: Johns Hopkins University flagged NASA’s Artemis II crewed lunar flyby, expected in early 2026, as the year’s marquee space story, the kind of scheduled milestone that alone could nudge a market like this toward resolution. But the structure is what makes it interesting to a prediction-market audience: a near-certain outcome forces traders to reason about exactly which triggers have fired and when, rather than whether the headline claim is plausible. That is the same logic behind more conventional markets we have covered, such as the multi-outcome betting around Spider-Man: Brand New Day.
What the 96% actually implies
As of August 01, 2026, the market sits at a 96% implied probability of YES on Manifold. In prediction-market terms, that price is the crowd’s best estimate that the resolution condition has already been met or will be met before the market closes on December 31. A 96% reading is not a coin flip; it is the market treating a YES outcome as close to settled while leaving a small residual for disputes over interpretation.
One caveat matters for finance readers. Manifold runs primarily on play money, its in-platform currency called mana, rather than regulated cash contracts. So the “money” here is a measure of conviction and reputation, not dollars at risk. That distinguishes it from cash exchanges such as Kalshi, the CFTC-regulated US venue, and from Polymarket, which settles in stablecoin and is not available to US persons. The signal is still useful, because Manifold traders build track records, but it should be read as a sentiment gauge rather than a regulated financial instrument.
Why the market moved this week
The catalyst is a genuine mathematics story. On July 20 and 21, 2026, Levent Alpöge, a mathematician affiliated with Harvard University and the AI company Anthropic, announced a counterexample that disproves the Jacobian conjecture, a problem posed in its general form by Ott-Heinrich Keller in 1939. Alpöge credited Anthropic’s Claude Fable 5 model with finding the object, and posted the result on X during the World Cup final. According to Fortune, the post drew more than 20 million views.
The Jacobian conjecture asks whether a polynomial map with a nonzero constant Jacobian determinant must be reversible with a polynomial inverse. The counterexample Alpöge published is short enough to fit in a single social-media post: as The Conversation reported, it describes a three-dimensional map whose Jacobian determinant holds steady at negative two everywhere, yet sends three distinct starting points to the same destination, which means it cannot be reversible. The result disproves the conjecture for dimensions greater than two. The original two-dimensional case remains open.
Who and what is involved
The human at the center is Levent Alpöge, who works at the intersection of research mathematics and Anthropic’s AI effort. The tool is Claude Fable 5, an Anthropic model that Alpöge used to search for the counterexample, released into a crowded field of rival systems: NVIDIA disclosed in December 2025 that its own Nemotron 3 lineup would add a 49-billion-parameter Super model and a 253-billion-parameter Ultra model, both targeted for H1 2026. Independent verification came quickly: mathematician Kevin Buzzard of Imperial College London confirmed that the example checked out in the Lean proof assistant, telling Fortune, “It is a big day. I think it’s a great time to be alive, personally.”
Not everyone framed it as celebration. Akhil Mathew of the University of Chicago described the episode as “a very rapid and very unsettling change,” noting that the AI supplied the “how” without the “why.” That tension, between a verified result and an uneasy sense of how it was produced, is part of what has drawn attention back to a market that was designed as a throwaway joke. For readers tracking the underlying models, it fits the broader arc of Anthropic’s year, including the way Claude Opus 4.8 topped an AI leaderboard in 2026, even as NVIDIA counters with its own roadmap: the smaller Nemotron Nano 3 models are already shipping, with the larger Super and Ultra versions due by June 2026, according to NVIDIA’s own December 2025 disclosure.
The resolution fight driving the odds
Here is where the novelty market gets its edge. One of the listed YES triggers is an AI solving a famous unsolved mathematics problem, and the Jacobian conjecture disproof lands squarely in that gray zone. Comments on the market debate whether the result should count as an AI directly solving the problem, or whether it is better read as a human mathematician using an AI as a tool. The market creator has signaled a strong lean toward resolving YES, but the final call appeared pending as of early August.
That debate is the whole game at 96%. The remaining few percentage points are not a bet that nothing will happen in 2026. They are a bet on definitional risk: whether the resolver treats the Jacobian result, or one of the market’s many other triggers, as clearly qualifying. It is a small, specific disagreement wearing the costume of a cosmic question.
A 2026 timeline of AI math milestones
The Jacobian result did not arrive in isolation. It caps a run of AI-assisted mathematics that has compounded over roughly a year, which helps explain why traders treat the “something will happen” bar as effectively cleared.
| Date | Milestone | Why it matters |
|---|---|---|
| Mid-2025 | AI models solve five of six problems at the International Mathematical Olympiad | First broad signal that models could handle competition-grade proofs |
| May 2026 | OpenAI reports disproving an 80-year-old Erdős conjecture in combinatorial geometry | Shifted the conversation from contest problems to open research questions |
| June 2026 | Researchers publish the Leiden Declaration on AI and Mathematics | Called for guardrails on transparency, attribution, and peer review |
| July 20-21, 2026 | Alpöge and Claude Fable 5 disprove the Jacobian conjecture for dimensions above two | Verified in Lean within a day; described as the biggest AI-assisted result yet |
| August 01, 2026 | “Will something happen in 2026?” trades at 96% YES on Manifold | Traders debate whether the math result triggers resolution |
Why a joke market matters to tech and finance readers
Novelty markets like this one are more than internet humor. They stress-test how prediction markets handle ambiguity, resolution authority, and the gap between a headline and a precise rule. For anyone using markets as a forecasting tool, the “Will something happen in 2026?” case is a clean illustration of resolution risk: the probability of the underlying event can be near certain while the payout still hinges on a judgment call. That same dynamic shows up in serious cash markets, including the Kalshi odds on Trump’s next attorney general, where the wording of the resolution criteria can matter as much as the news.
There is also a signal-quality lesson. The 24-hour trading surge here, on the order of 150 to 170 active traders by early August, is a reminder that volume spikes on Manifold often track a specific external event rather than a change in the fundamental question. Reading the comments, not just the price, is how you find the actual catalyst. The Manifold odds swings around Spider-Man: Brand New Day followed the same pattern of price movement chasing discrete news.
What to watch next
Three things will decide how this closes. First, the resolver’s ruling on whether the Jacobian disproof qualifies as an AI solving a famous unsolved problem; a YES call would likely settle the market well before December. Second, peer review: the counterexample was verified computationally in Lean, but the underlying work is not yet journal peer-reviewed, and mathematicians are still absorbing it. Third, the two-dimensional case of the conjecture remains open, so the story of AI-assisted mathematics is not finished even for this single problem. For the market itself, the practical question is not whether something will happen in 2026, but which trigger the resolver formally recognizes first.
Frequently asked questions
What does “Will something happen in 2026?” actually resolve on?
It resolves YES if any item on a long list of defined triggers occurs, ranging from major political changes and market crashes to an AI solving a famous unsolved mathematics problem. It is a joke premise attached to a concrete rulebook.
Is real money at stake?
Manifold runs primarily on play money called mana, so the price reflects conviction and reputation rather than cash at risk. That is different from regulated venues like Kalshi or stablecoin-settled Polymarket.
Did an AI really disprove an 87-year-old math problem?
Yes. On July 20-21, 2026, Levent Alpöge announced a counterexample to the Jacobian conjecture found with Anthropic’s Claude Fable 5. It was verified in the Lean proof assistant within a day and disproves the conjecture for dimensions above two.
Why is the market at 96% and not 100%?
The residual reflects definitional risk, mainly whether the resolver treats the math result, or another trigger, as clearly qualifying, rather than any real doubt that events will occur in 2026.
The Bottom Line
Sources
- Manifold — Will something happen in 2026?
- Fortune — Mathematicians grapple with a very rapid and very unsettling change as AI cracks the Jacobian conjecture
- The Conversation — Why a tiny social media post has mathematicians rethinking AI
- The Week — The Jacobian conjecture disproved by AI, explained
- CoinDesk — Claude’s Fable 5 just solved an 87-year-old math problem
- TechCrunch — OpenAI claims it solved an 80-year-old math problem
Prediction markets carry risk and are not investment or betting advice. Availability is restricted by jurisdiction: Polymarket is not available to US persons, and Kalshi is a CFTC-regulated US exchange. Manifold runs primarily on play money. Participation is limited to those 18+ or 21+ as applicable in your area. If gambling is a problem for you or someone you know, call 1-800-GAMBLER for confidential help.
