The Trillion-Dollar AI Bet and Who Owns Your Ideas (Part 1)

From autocomplete to frontier intelligence

Not that long ago, you were typing an email and the app suggested what you were going to type next. You may have wondered: “How did it do that?” At the time, very few people realized that the ability to make those suggestions would create the world we live in today: a world of frontier intelligence and AI agents that can reason.

Those who saw the power of the technology made big bets on the future. Trillion-dollar bets. The kind of bets you can’t unwind.

Fast forward to this month, and a frontier AI company has announced that it solved the Clay Millennium Prize problem for the Navier-Stokes equation. If you aren’t a mathematician, think of these problems as the pinnacle of human intelligence. Solving just one changes the course of history and nets you a $1 million prize. It’s exactly the kind of win a trillion-dollar bet needs to survive.

Why the breakthrough isn’t the headline

Normally, I’d be celebrating the math and trying to explain to people who didn’t care that the Navier-Stokes equation can blow up in finite time. In any normal world, the breakthrough would be the headline. Instead, I’m writing about trust in the age of frontier intelligence.

The proof the AI found wasn’t a clean line from existing knowledge to something new. It detoured through the work of a couple of researchers who had used that same frontier AI to think through new ideas. From a distance, their ideas look like a simpler version of the bigger problem they were attacking. Take their work, add viscosity back into the fluid equations, and you’re roughly where you need to be to claim the $1 million prize after a lifetime of dedication. The frontier AI lab with the trillion-dollar bet had other plans.

How the proof actually got found

When the lab learned a solution to Navier-Stokes was near, it let loose an army of agents on that specific problem, running an unreleased version of its model. And just like that email you wrote all those years ago, the agents reasoned and predicted the next logical idea from decades of human research. They may have done it using the chat sessions of the people closest to the answer.

However you split the credit between real people and AI, one thing is now obvious: a centralized AI that takes your ideas can use them to capture the value you’re building.

Your ideas aren’t safe in someone else’s model

Even if a frontier lab isn’t going after your work directly, I don’t think anyone can argue that the ideas you share with frontier AI are safe anymore. It’s not that the lab will take your ideas specifically. It’s that they’ll improve their models with them. Then someone out there will take them. And by someone, I mean your competition.

The irony isn’t lost on me. The trophy that turns the trillion-dollar bet into a winning bet is the same moment everyone needs to take control of their own AI knowledge and reasoning. Luckily, open-weight models make that possible.

What’s next

In the next post, I’ll look at a recently released model you can run locally on modest hardware. It has exceptional reasoning ability and no plans to steal the $1 million cake out of your oven. Then we’ll look at a new way to control its reasoning effort to get the most value out of it.

About the author

Rafael “Raf” Rubio is an HPC Engineer at Hoonify Technologies in Albuquerque. He works on the high-performance computing systems behind Hoonify’s inference platform, which runs open models on TurbOS so customers can use AI they own and control. Raf studied at New Mexico State University and is part of the New Mexico engineering team that has deep roots in the state’s national lab community.