Key Takeaways

  • Vlad Tenev co-founded Harmonic to pursue mathematical superintelligence; its Aristotle model was one of three systems to earn a gold medal on the International Math Olympiad last year.
  • Leading artificial intelligence labs are centralizing scientific research by running secret models inside private server clusters, attempting to monopolize drug discovery and scientific breakthroughs internally.
  • Tenev argues that hoarding scientific discoveries inside corporate clusters is socially unacceptable and will lose to models distributed directly to end users.
  • Distributing formal verification tools like Lean to working mathematicians creates verifiable truth while avoiding public backlash against automated research silos.

The Closed Lab Monopoly on Science Will Fail

Large artificial intelligence labs have settled on a playbook: build massive computing clusters, train proprietary models behind closed doors, and capture the upside of every scientific breakthrough internally. Tenev sees this path as dangerous and fragile. When frontier labs use proprietary clusters to monopolize mathematics, chemistry, and drug discovery, they separate technological progress from the people who understand the field.

“The one thing I don't love about like this latest wave of AI models solving math problems is that it's too centralized,” (play from 14:48) Tenev told Sarah Guo. “It's like AI companies with their big data centers and their employees using secret models to sort of do all this work.” (play from 15:00)

The end state of that strategy is a closed loop where a tiny group of tech executives controls scientific patents, medical treatments, and intellectual property. Tenev calls that outcome unrealistic: “If all of that is happening and it's owned by the AI labs themselves, I think that's a bit of like a dystopian outcome and unlikely to happen.” (play from 17:00) Society will push back hard against any single company trying to lock up scientific discovery inside private servers.

Formal Proofs Work Best in the Open

Last year, Harmonic's Aristotle model proved that automated systems could compete with elite human problem solvers, becoming one of three models to earn a gold medal on the International Math Olympiad. But for Tenev, building Aristotle was not about creating an internal research moat. The goal is putting rigorous mathematical tools into the hands of the broader research community.

In formal mathematics, software environments like Lean allow proofs to be verified by machine code. When an answer is correct, the compiler verifies it beyond doubt. That removes the hallucination problem that plagues standard text models. Because proof verification is objective, progress does not require trust in a single tech giant. It requires giving the verification software to independent researchers so they can test their own ideas.

“What would ultimately win out in the end as a model is rolling these things out and actually putting them in the hands of people and having the users make interesting discoveries and inventions,” (play from 15:13) Tenev explained. Instead of treating researchers as legacy labor to replace, software builders should treat them as collaborators. “I think it would have been much better if they found some mathematicians and kind of collaborated with them to not make it seem like there's this dark data center dystopian future where one company is going to make all of the scientific discoveries.” (play from 18:32)

What to Do With This

Audit your product roadmap this week. If your system automates high-stakes work, stop trying to run the entire domain inside a black-box service that replaces human practitioners. Build a verifiable interface, like a compiler or a formal proof checker, and place the tool directly in front of three working practitioners in your target industry by Friday. Let external domain experts generate the discoveries while your software provides the verification engine.