Cogentic uses a multi-agent prove-verify loop to produce expert-verified new results on open problems in online learning, auction theory, and mechanism design
The authors present Cogentic, a multi-agent harness for automated proof discovery on open research problems: an orchestrator allocates a population of independent provers across distinct proof directions, several specialized components subject their output to adversarial verification, and confirmed intermediate results are promoted into a persistent verified ledger that later rounds build on; using either Gemini 3.1 Pro or an early version of Gemini 4 Argon as the base model, the harness produced novel results on open problems across online learning, auction theory, and mechanism design, each independently verified by domain experts and developed in full in companion papers.