Who we are

We invented language to share reasoning, and math to make it exact. Then we built machines that learn from everything we'd written. Same pursuit each time: understand the world a little better. The machines learned from us. Now it runs the other way. Every day, more people ask a machine to think for them.

They've reached the frontier. In July 2026 a mathematician working with a frontier model found a counterexample to the Jacobian conjecture, open since 1939. Terence Tao told the International Congress of Mathematicians that month that the field is entering "a crisis in the foundations of mathematical values and practices."

We scaled pre-training, added the right reinforcement learning, and models helped overturn conjectures that had stood since the 1930s and 40s. Those wins came from one thing: a reward the model couldn't fake. A proof either checks or it doesn't. Soon they'll produce proofs a machine can check but no person can explain. The next scaling law is recursive self-improvement, models that make the next model better, and it needs a second reward like that: fast, real, and impossible to fake. It already exists.

The financial markets. A decision gets a price in seconds, and the price is real money.

Every trade has a winner and a loser, and the market marks which is which in seconds. A model that learns here is graded on every action, in money. It's a fast reward and a noisy one, and telling skill from luck is the whole job. With machines in the loop, the winners take more. We help people find their edge and press it. As the machines improve, so do the best humans.

We don't believe in replacing people. We believe in augmenting them. We don't build models. We take the strongest ones and point them at the market, so a trader can find the signal in the noise. A probabilistic model makes mistakes, and here a mistake is the difference between making money and losing your entire balance. So we built the harness that closes that gap: the checks and limits a model needs to operate where an error costs real money. The person stays in charge: they see the case for the trade and can overrule it.

We built the only trading platform that gets better with you. GTE is live today. That's just the beginning.

Alea iacta est. - Liquid Labs