Kristian Fagerlie builds an AI research workflow on a virtual server using Codex and a browser agent. The system watches selected social posts, compares them with prediction-market questions and scores relevance, directional impact and novelty, while filtering repeated information.
Kristian Fagerlie demonstrates the research dashboard and compares a reported post timestamp with a later market-price movement. The example suggests a possible information lag, but it does not establish that an executable trade was available at the displayed price or that the workflow would be profitable.
Kristian Fagerlie emphasizes limitations in sampling, latency and liquidity. The system remains a shadow-testing research experiment, not a trained probability model or demonstrated autonomous trading strategy. The account does not provide financial advice or treat one timing example as a repeatable advantage.
Watch on YouTube
