Socratica

Cosmos Institute granted Jasmine Cui on August 15, 2025 for Socratica, with funding in the $1K to $10K range plus compute. The grant came through the inaugural AI x Truth-Seeking cohort of 27 winners co-funded by Cosmos Institute and FIRE under a $1M program.

Cui is an NBC News reporter and data journalist, a fellow at MIT's Algorithmic Alignment Group, and a MATS scholar working on mechanistic interpretability. She holds a degree from Columbia Journalism School and worked previously as a data scientist across accounting and telecommunications. Her hybrid background (newsroom data work plus interpretability research) is the unusual combination that the cohort's editorial direction tends to favor.

Socratica is positioned as an AI librarian that conducts research and helps users formulate better questions, treating question-quality as a first-class output rather than a means to an answer. The product sits closer to a research-coaching tool than to a search interface, and the design assumption is that for many inquiry tasks the bottleneck is the question, not the retrieval. The 90-day Cosmos x FIRE prototype window applies.

Within the open AI stack Socratica sits in the agents and retrieval-memory layers. Within the cohort it is one of three Socratic-method projects (alongside Socratic Mirror and Argument Debugger), a thematic cluster distinct from the model-probing tools (Perplex, Reward Hacking Benchmark) in the same cohort. The lock-in vector is similar to Socratic Mirror's: commercial assistants optimize for answer delivery, which makes question-formulation a neglected capability surface.