Harmonic, the artificial intelligence startup co-founded by Robinhood CEO Vlad Tenev, announced the beta release of a chatbot app for iOS and Android, through which users can interact with the company’s AI model, Aristotle.
Harmonic aims to broaden access to Aristotle, which claims to provide answers free of “hallucinations” (misleading or false information) for queries involving mathematical reasoning — a particularly ambitious undertaking given the general reliability challenges of today’s AI models. The company’s vision is to create a “mathematical superintelligence” (MSI) that will assist users in all fields relying on mathematics, such as physics, statistics, and computer science.
The co-founder and CEO of Harmonic, Tudor Achim, told TechCrunch that Aristotle is the first product available to the public that performs logical reasoning and formally verifies its outputs. The system supports domains of quantitative reasoning and guarantees the absence of erroneous or false answers within those areas.
The company also plans to release an API for enterprises and a web app for consumers. Aristotle earned a gold medal at the 2025 International Mathematical Olympiad (IMO) through a formal test, where problems were translated into a machine-readable format. AI models from Google and OpenAI also achieved gold medals at this year’s IMO but via more informal tests in natural language.
Harmonic used the open-source programming language Lean to generate responses, and before providing the final answer to users, the system double-checks correctness through an algorithmic process that does not involve AI. This method is similar to verification systems used in critical fields such as medical devices and aerospace engineering.
The beta app release comes just weeks after Harmonic raised $100 million in a Series B funding round led by Kleiner Perkins at an $875 million valuation. The CEO states that the company is rapidly advancing toward achieving mathematical superintelligence, and the large funding reflects the ambitious scope.
Achieving reliable answers without hallucinations in an AI model is extremely challenging, as even the top AI solutions still exhibit hallucinations, and this problem does not seem to be improving with newer versions, as has been reported for OpenAI models.