Information Markets, Decision Markets, Attention Markets, Action Markets

A clean, generative taxonomy of prediction-market variants: information markets (predicting past/present via a trusted resolver), the conditional-market trick ('if Scott reviewed X, what would he find?' — resolve one at random, refund the rest, as in replication markets), decision markets (advise a decision-maker who resolves by acting), attention markets (the Aaronson-crackpot-filter: only read the manifestos the market scores highly), and action markets — generalizing the assassination-market worry into the elegant law that 'every prediction market is also an action market, and vice versa,' with a leakage spectrum (supernova ~100% prediction, presidential-death ~99%, backyard-cleanup ~80% action). The closing move connects this to AI safety (a predictor can cause the event it predicts) and, beautifully, to the active-inference theory of motor control (the brain predicts your arm will move at 100%, then moves it to 'win the bet'). Memorable and unusually fertile.

Quality 74 · Strong. Strong, near-Excellent: an original, memorable framework with a genuine unifying insight (prediction↔action duality) that pays off across mechanism design, AI safety, and neuroscience; a standout conceptual essay held just below Excellent by its exploratory, list-like structure.

Claude’s paradigm shift 54 · Moderate. Notable-major: the 'every prediction market is also an action market' law with the leakage spectrum, and the active-inference tie-in, are a fresh and generative synthesis.

Real-world impact 3 · Moderate. Moderate (3): the conditional-/decision-/attention-market designs are concrete mechanism-design ideas with real traction in the forecasting/EA community.