Show HN: OpenThomas – weather trader agent for prediction markets (github.com)

OpenThomas, an autonomous AI trading agent, has been launched to engage in weather prediction markets on platforms like Kalshi and Polymarket. This innovative tool builds its trading strategies using a consensus from seven weather models, adjusting for unique biases learned from extensive historical data. The agent operates within strict risk limits by employing fractional Kelly sizing, allowing it to make informed trades without the need for wallet integrations or API keys initially, making it accessible for new users. The system emphasizes a disciplined approach to trading, using a framework that prioritizes measurable and learnable advantages in real-time, ultimately aiming for steady growth while safeguarding capital. The significance of OpenThomas lies in its structured methodology, which contrasts with the general lack of profitability seen among other AI models in similar environments. The model not only leverages systematic biases in weather forecasts but also incorporates sophisticated decision-making layers that filter markets for potential mispricings. By calibrating its forecasting against actual market behaviors and learning from each trade's outcomes, OpenThomas presents a unique intersection of AI and finance. This represents a potential breakthrough for the AI/ML community, demonstrating that disciplined trading strategies, grounded in data-driven insights, can navigate the complexities of prediction markets more effectively than traditional pure forecasting models, which tend to underperform.