One architecture, applied where it has an advantage today and grown toward where we are going.
Quantum simulation produces states that are read into a few numbers and discarded. Our models take the states themselves, learn the rule that governs them, and carry it where the source cannot go.
The same architecture, grown across the processor. Quantum data is where it proves itself first, and what it learns there is what takes it the rest of the way.
A simulation hands you outcomes, never the rules underneath. Turning one into the other is work that today takes a senior physicist weeks and carries no guarantee.
Learned on a corner of a system, run across the whole of it.
Any horizon answered by a circuit whose size does not grow with it.
The model certifies its own answer, and refuses when it should.
Ask what a variant does without returning to the source.
We would like to hear about the problem.