AI that keeps learning
after it ships.
Evi, Kus and Revmosys run on the same kind of AI everyone else builds on. Alongside them, we're researching a different approach: an AI modeled on how brain cells learn, designed to keep learning after launch.
Today's AI is built on one specific idea.
Large language models, including the ones behind our own products, are trained once on enormous datasets with enormous compute. When training stops, the model is frozen: it doesn't keep learning from what happens after it ships.
An AI modeled on how brain cells learn.
We're building an AI modeled on how brain cells learn, designed to keep learning from what it encounters after launch.
We'll publish the specifics once the results earn it. For now, this page describes the direction.
A different bet on how learning should work.
What's real today, plainly stated.
This is active research. Today the system can take in real data and organize what it has seen. It can't yet solve new problems, generate answers, or do anything a user would see, and we'll report each of those steps here when it happens.
A long-term bet, said plainly as one.
Scaling today's approach is one path to more general intelligence, and we're betting on another. A system that keeps learning from experience, the way brains do, could be a necessary ingredient for it. That bet is speculative and early, and it may not work.
Following this closely?
We'll share results here as they happen. For anything else, write to us directly.