AI Milestones 2026 — AGI, Millennium Prize and releases

Kalshi's two OpenAI release ladders—GPT-6 and Astra—have both settled, and the boards that replace them on this page are Polymarket's, which are smaller and worse behaved.

Polymarket's Sol (5.7+) ladder prices by September 30 at 83.0%, by October 31 at 91.5% and by November 30 at 94.0%. That slope is at least monotonic. Its Astra (6.1+) ladder is not: by October 31 is 47.5% and by December 31 is 85.5%, on a board summing to 133.0% with $349 of daily volume—two rungs, one price, and nobody trading it. Between them the two ladders turn under a thousand dollars a day.

The Manifold anchor is where the real money is. The Navier-Stokes contract—will Anthropic announce a solution—sits at 12.2% and has taken $21,267 today against $256,213 of lifetime volume, making it the largest single contract on the exchange. It settles November 1.

The multi-year boards frame it. Kalshi's Millennium Prize board: before 2028 18.9%, before 2029 44.0%, before 2030 66.1%, before 2035 70.6%, on $14,726—and the flow is concentrated at the near end, $7,455 on the 2028 rung. The AGI boards are the tightest pair here: Kalshi's OpenAI announces AGI before 2027 at 24.9% against before 2028 at 61.3%, on $14,162 and $738,816 of lifetime volume.

The any-company AGI ladder sums to 747.5% across 12 rungs and quotes before 2031 at 89.5% with zero trades, while before Oct 2028 at 87.0% took $476. The long end of every board on this page is quoted and untouched; all the money sits inside three years.

The Manifold anchor and its annual sibling. The Navier-Stokes contract settles November 1 and asks a narrower question than the others: not whether an AI solves the problem, but whether Anthropic announces it, before an IPO. At 16.5% on $176,497, it is the most heavily traded contract on that exchange today.

The one-month question against the multi-year one. Manifold prices an AI solving a Millennium problem this month at 10.3%; Kalshi's ladder puts the next prize being awarded before 2027 at 5.05% and before 2029 at 41.9%. Solving and awarding are different events separated by the Clay Institute's two-year publication requirement, which is most of the gap—but not obviously all of it.

AGI on two exchanges, two points apart. Polymarket asks whether OpenAI announces it; Kalshi asks whether OpenAI achieves it. Both resolve on public statements in practice. Kalshi's ladder runs to 2030 and holds $345,884, most of it on the near legs.

Polymarket's two OpenAI release ladders, standing in for the Kalshi GPT-6 and Astra boards that settled. Sol (5.7+): September 30 at 83.0%, October 31 at 91.5%, November 30 at 94.0%, on $621—properly sloped. Astra (6.1+): October 31 at 47.5%, December 31 at 85.5%, summing to 133.0% on $349. Two rungs is not a ladder, and the pair together turn less than a thousand dollars a day.

Polymarket's Luna ladder against Kalshi's any-company AGI board. The Kalshi board sums to 747.5% across 12 rungs—cumulative, so that is structural—but its longest rungs are quoted without trading: before 2031 at 89.5% on , before 2030 at 86.5% on , while before October 2028 at 87.0% took $476. Every dollar on this board is inside three years, and every quote beyond that is a placeholder.

The Navier-Stokes contract settles November 1 and at 12.2% it is both the largest single market on Manifold and the sharpest disagreement with the multi-year Millennium boards, which price any AI solving any of the seven problems before 2028 at 18.9%. One lab, one problem, two months—against seven problems, every lab, two years. The gap is six and a half points.

Both Kalshi release ladders have settled and Polymarket's replacements are thin. The Astra ladder's two rungs at 47.5% and 85.5% on $349 is not a market anyone should read as a forecast. Watch whether Kalshi relists.

The AGI pair is the best-priced thing on this page—24.9% before 2027 against 61.3% before 2028, on $738,816 of lifetime volume. That is a real term structure and it moves.

Everything beyond three years is quoted, not traded. The any-company ladder's 2030 and 2031 rungs have taken nothing at all today, and the Millennium board's 2035 rung likewise. The tradeable horizon on AI milestones is currently about 24 months.