Descriptions:
The All About AI channel published a detailed first-look at using Anthropic’s Claude (operating through Claude Code) as a fully autonomous agentic trading system on Polymarket’s 5-minute up/down prediction markets. The creator fed 24 hours of collected market data to the model and asked it to analyze, design, implement, and execute a positive-expected-value trading strategy from scratch.
Claude produced a formula-based fair-value strategy that incorporates Binance price feeds to detect arbitrage signals against the Polymarket order book, accounts for exchange fees in profitability calculations, and applies Kelly criterion position sizing at 25% of full Kelly with a 10% daily loss halt. A notable emergent behavior was the model independently developing a secondary “deep long shot fading” sub-strategy — deliberately taking low-probability positions where the potential payout substantially exceeds the break-even probability implied by the model’s own fair-value estimate. The creator configured cron-job-based automated monitoring directly within Claude Code, enabling the agent to wake every two hours overnight, audit recent trades, adjust strategy parameters if needed, and restart the bot — all without human involvement.
After approximately 15 hours of live operation, the system reported a 71% win rate and roughly $41 in net gains. The video is a practical end-to-end walkthrough of autonomous agentic workflows inside Claude Code, covering prompt design, iterative mid-session strategy review, and self-healing monitoring loops — useful for developers exploring agent autonomy for data-driven or financial applications.
📺 Source: All About AI · Published June 10, 2026
🏷️ Format: Showcase
