Prediction
Answer Options
As of 11 October 2026, 00:28 UTC, versus prices Yes at a 3% chance and No at 97%.
YES if, by 31 October 2026 at 11:59 PM ET, OpenAI or Anthropic (the company itself, or someone officially speaking on its behalf) announces that it has achieved recursive self-improvement; otherwise NO. Here, recursive self-improvement is the capability, as set out in the company's announcement, where the company's AI by itself carries out the research and development that materially raises the general capabilities of the company's most capable AI systems, or of its near-frontier AI systems, or that produces a successor more capable than them, and where the improved system is in turn able to drive further improvements of that kind. Humans may define high-level aims and give resources, supervision and sign-off, but the substantive research decisions and the technical solution must not come from them. Toy, narrow or sandboxed improvements, proofs of concept, and gains that only touch prompts, skills or benchmark scores do not count. The announcement must be made in an official capacity and must describe the capability as achieved at present. None of these count: statements that AI helps, speeds up or writes code for the company's research; plans, expectations or forecasts; speculation; characterisations by interviewers; leaks; model outputs; and claims from third parties. The company does not need to release anything publicly, name a model, give a technical demo or obtain independent verification. Official information from OpenAI or Anthropic, or from their official representatives, decides; broad agreement in reputable reporting may also be used.
Each percentage is the market’s current estimate of how likely that answer is. A contract settles at 100p if the answer is right and at 0p if it is wrong.
Reported by versus Markets
