Exploiting information asymmetry in the stock market with OpenAI’s flagship model, GPT-5.2
A breakdown of our recursive strategy to uncover data divergences in financial filings, including the prompt set and a new stock recommendation.
📶BUY $PSN (Parsons Corp) and hold for 22 days till February 18th, 2026.
After two strong runs with this Deep Research strategy, we ran it again using OpenAI’s flagship model, GPT-5.2, in ChatGPT’s Deep Research mode. This time, we asked GPT-5.2 to recursively improve the prompts from the last two runs, with the goal of increasing the ROI of the next stock recommendation.
The objective was to find mispriced stocks: companies where primary-source financials show improving fundamentals, but the share price hasn’t caught up yet. GPT-5.2 scanned filings and earnings materials, stress-tested the ideas, and anchored each pick to a dated catalyst and a defined exit plan. This memo walks through the full workflow and shares the single highest-conviction long it produced.