Harmonic, CEO of Robinhood, AI Math Startup, launches the AI Chatbot app

Harmonic, an AI startup co-founded by Robinhood CEO Vlad Tenev, announced on Monday it will be beta launch of the iOS and Android chatbot app, which will allow users to access AI model Aristotle.

With this launch, the company aims to expand access to Aristotle. Aristotle argues that harmonics provide a “haptic-free” answer to questions that contain mathematical inference. This is a bold claim taking into account the reliability issues of today’s AI models. Harmonic focuses on creating “mathematical tensions” or MSIs. The startup ultimately wants to help users in all areas of mathematics-dependent, including physics, statistics, and computer science.

“(Aristotle) is the first product available to people who make inferences and formally verify the output,” Harmonic CEO and co-founder Tudor Achim said in an interview with TechCrunch. “With the domain supported by Aristotle, it is a quantitative inference domain, so it ensures that there are no hallucinations in practice.”

Ultimately, Harmonic said it plans to release the API to give enterprises access to Aristotle and consumer web apps.

The Beta launch is weeks after raising $100 million at a $875 million valuation in the Series B round led by Kleiner Perkins. Achim claims that Harmonic is “moving very quickly” along the path to achieving MSI, and that investors believed it was a fair rating given the scope of his startup’s ambitions.

Achim says Harmonic achieves its overly accuratic solution by having Aristotle generate responses in the open source programming language Lean. Before Aristotle gives the user the answer, he says that the model double-checks the solution to be correct via an algorithmic process that includes AI. Harmonic CEO points out that similar technologies will be used to verify power output in high-stakes fields such as medical devices and aviation.

Even in narrow domains, achieving hallucination-free performance from AI models is a very difficult task. Research has discovered that Even major AI models hallucinate a lot of hallucinationsand the problem doesn’t seem to be getting better. Openai’s latest AI inference model is more hallucinating than older ones.

Harmonic says Aristotle achieved gold medal performance at the 2025 International Mathematics Olympiad through formal tests (meaning that the problem was converted to a machine-readable format). Google and Openai also We developed an AI model that achieved gold medal performance at this year’s IMObut through unofficial tests filmed in natural language.