We are building infrastructure for a world where intelligence is abundant and confidence is scarce. The work is under wraps. The conviction is not.
The interesting question is no longer whether a machine can produce the work. It is whether anyone can tell that the work is right. Those are different problems, and only one of them is getting solved.
Producing the work is close to solved. Deciding whether the work is any good is not, and the gap between those two widens every month.
The failure that costs money is not obvious nonsense. It is output that reads exactly like the right answer, is not, and goes unnoticed until somebody has already acted on it.
Sounding certain is free. Being right is not. Somebody has to build the difference between the two, and remarkably few people are working on it.
Anyone can produce the work now. The open question is who can tell you it is right.
We would rather show the working system than the pitch for it. Until then the thesis is public, the roadmap is not, and the fastest way in is to write to us.
Privately funded. Small team. Long horizon.
We will write when there is something worth showing, and not before.
One address. No list, no newsletter. We write once, when there is something.