Harmonic AI, the formal reasoning startup co-founded by Robinhood’s Vlad Tenev and Tudor Achim, is making a pointed argument in the AI industry: the benchmarks we use to measure artificial intelligence should be designed by humans, not generated by other AI systems.
The company’s stance matters beyond the AI research bubble. As crypto and DeFi protocols increasingly embed AI agents into trading systems, risk models, and smart contract auditing tools, the question of how we verify that AI actually works correctly is no longer academic.
What Harmonic is actually building
Harmonic’s core product is Aristotle, a formal reasoning system designed to produce mathematically verifiable outputs. Think of it as the difference between an AI that says “this code probably works” and one that can prove it does, the way a mathematician proves a theorem.
Harmonic recorded gold medal-level performance at the 2025 International Mathematical Olympiad. The system also scored 96.8% on the VERINA code verification benchmark in December 2025.
