Harmonic: AI Math Startup Founded by Robinhood’s CEO Launches Innovative AI Chatbot App

Harmonic, an AI startup co-founded by Vlad Tenev, CEO of Robinhood, announced on Monday the beta launch of a chatbot app for iOS and Android, enabling users to engage with its AI model, Aristotle.

Through this release, the company aims to broaden access to Aristotle, which Harmonic claims delivers “hallucination-free” answers for questions involving mathematical reasoning. This assertion is notable, given the reliability challenges faced by current AI models. Harmonic’s goal is to create “mathematical superintelligence” (MSI) to support users in fields reliant on mathematics, including physics, statistics, and computer science.

“[Aristotle] is the first product that enables reasoning and formally verifies its results,” said Tudor Achim, CEO and co-founder of Harmonic, in a discussion with TechCrunch. “In the quantitative reasoning domain covered by Aristotle—we can guarantee no hallucinations.”

Looking forward, Harmonic intends to launch an API for companies to utilize Aristotle, along with a web application for everyday consumers.

This beta introduction follows Harmonic’s recent $100 million Series B funding round led by Kleiner Perkins, which has given the startup a valuation of $875 million. Achim emphasizes that Harmonic is “moving quickly” towards realizing MSI, and that investors considered this valuation fitting given the startup’s ambitious aspirations.

Achim elaborates that Harmonic guarantees hyper-accurate responses by having Aristotle formulate answers in the open-source programming language Lean. He notes that before delivering a response, the model verifies the accuracy of the solution using an algorithmic method independent of AI. Harmonic’s CEO points out that similar verification technology is used in critical fields such as medical devices and aviation.

Achieving hallucination-free performance in AI systems, even in specialized areas, is notoriously difficult. Research reveals that even top-tier AI models often generate hallucinations, and the situation appears stagnant. Newer versions of OpenAI’s reasoning models show more hallucinations than their predecessors.

Harmonic asserts that Aristotle has demonstrated gold medal performance at the 2025 International Math Olympiad through a formal evaluation (where problems were provided in a machine-readable format). In contrast, Google and OpenAI have also created AI models that achieved gold medal status in this year’s IMO but relied on informal assessments conducted in natural language.