Asks whether large language models can outperform prediction market consensus prices and argues the more tractable framing is using LLMs as updaters rather than predictors. Distinguishes cold prediction (generating a probability estimate without prior context) from updating (revising an existing estimate as new information arrives), and considers what each role implies for AI tools deployed alongside human traders in live markets.

Data-driven deep dive into Polymarket's order book structure using 600M+ raw datapoints filtered to a 343M research dataset. Categorizes order flow into soft (retail), hard (professional), and AI flow, revealing that Polymarket's liquidity is episodic and attention-driven: the p95 peak hour shows hundreds of millions in open interest while the p50 median is thin. Order book analysis shows surface symmetry at top-of-book but systematic ask-side skew at deeper levels, and market impact data confirms that medium-to-large orders hit liquidity cliffs. Argues the core problem is trapped capital — dollars reserved multiple times against mutually exclusive outcomes — and that better netting and capital efficiency, not more money, is the fix.

Comprehensive research report on AI agents for prediction markets, proposing a four-layer architecture (data, analysis, execution, learning) for autonomous trading systems. Maps the ecosystem of existing agents, compares Kelly criterion vs fixed-fraction bet sizing, surveys arbitrage strategies across platforms, and outlines business models (agent-as-a-service, liquidity mining, data sales). Argues that AI agents will become the dominant market participants within two years, transforming prediction markets from retail-driven speculation into infrastructure for probabilistic information.

Argues prediction markets are the natural marketplace for sovereign AI agents to trade their core commodity: information. Frames decentralized PMs as the 'bazaar' where agents monetize alpha through positions, market creators earn fees from surfacing unanswered questions, and reproducible computation enables incorruptible AI judges for dispute resolution. Positions this as an alternative to centralized AI lab alignment—market incentives align agents through financial participation rather than top-down instruction.

Six-month empirical analysis of political prediction market quality across Kalshi and Polymarket. Finds only 1.3% of political markets are liquid enough to be manipulation-resistant, bid-ask spreads exceed 20% on most contracts, and only 53% of resolved US elections appeared on both platforms. Proposes a four-part blueprint: stock relevant questions, cross-subsidize political liquidity from sports profits, deploy AI market makers where human interest is insufficient, and standardize contract definitions across platforms.

Examines early results from the Prediction Arena experiment, where six AI models trade real money on Kalshi. Five of six are underwater after three weeks, suggesting that raw information processing isn't enough to generate edge. Argues that successful prediction market traders profit from embodied, local knowledge (monitoring flights, calling embassies) rather than synthesizing public information, a domain where AI remains fundamentally constrained.

Argues the hardest PM problem isn't pricing but deciding what actually happened. Proposes cryptographically-committed LLMs as resolution judges—trading human bias and conflicts of interest for more tractable technical vulnerabilities. Cites Polymarket disputes (Venezuela election, Ukraine map, government shutdown) as evidence current systems fail at scale.

Galaxy research report on prediction markets evolving from niche speculation to mainstream financial infrastructure. Covers Polymarket ($9B valuation, 1.6M users) and Kalshi (top finance iOS app), emerging leverage mechanisms (Space, Gondor), AI as interface layer for fragmented venues, and convergence toward derivatives—event contracts as hedges, collateral, and composable financial primitives.

Proposes using AI as the linchpin to scale prediction markets to billions of users. Envisions an AI quartet: content creators (generating markets), event recommenders (personalization), liquidity allocators, and information aggregators. Argues this enables prediction markets at microscopic scale, making them personally relevant.