“An internal benchmark for OpenAI's Deep Research agent was to find all 11 papers co-authored by Liam Feddes and Barrett Zoff, a task the model can now mostly or fully complete.”