AI builds better AI for high‑GDP open worlds.
The open, high-stakes systems the economy runs on, where no test suite grades the answer. We build AI that improves itself there.
Read the thesisCode comes with a grader. The world does not.
In code and math, the interpreter says what happens next and the test suite says what it was worth. That is why almost all self-improving AI is built there.
The systems the economy runs on hand you neither. Take a marketplace, a supply chain, or a power grid: outcomes arrive months late, everyone else reacts to every decision, and the benchmark goes stale.
So we build the world and its grader ourselves, prove they can be trusted, and let AI improve itself inside them. That is most of the problem, and it is the part we work on.