Show HN: Reddit Stock Chatter Is Best Shorted (morethanamachine.com)

A new exploratory project called StockBot has emerged from an analysis of Reddit stock chatter over roughly 100 days, leveraging a bot that monitored discussions across over 10 subreddits. The project aimed to discover how a language model (Claude) would interpret noisy market data by assessing bursts of attention on specific tickers without directly integrating with trading platforms. The archive contained over 717,205 mentions, revealing that the median stock price fell 7.3% over ten sessions following these attention spikes. Importantly, the research highlights that while attention signals were informative, they should not be construed as trading directions; rather, they served as alerts requiring further investigation. The significance of this study lies in its examination of how model-driven insights can transform unstructured market chatter into actionable data, albeit with a cautionary note about their reliability. The bot employed different paths for generating stock alerts—systematic briefings and user-driven deep dives—each revealing different aspects of Reddit sentiment and its potential predictive power. The findings suggest that a simplistic trading strategy, like shorting every spike, would have yielded significant gains, although the system remains purely experimental and not indicative of a robust trading strategy. This project underlines the role of AI in deciphering market indicators from social media conversations, opening avenues for future research in sentiment-driven trading algorithms.