03 / A living field log
Still finding out.
These are keeping me up at night. Why do the best ideas always start arriving around midnight?
Last updated 29 Oct 2026
Research that can check itself
With CRRG, I’m exploring how an automated research process can keep a faithful record of what it has actually established. Can each proposed step be checked against the goal it is supposed to serve?
The mechanism is working in a live deployment. The next piece is going to be empirical validation.
Making room for useful surprises
The Organism is my ongoing experiment in giving agents room to explore while keeping formal verification at the boundary where claims become accepted results.
The next piece will add gazing and other features to solve the human <> agent communications bottleneck.
Following the bottleneck
The prover optimization series follows experiments from Plonky3 through LeanVM. There are many provers, but my next piece is fully freeroaming autoresearch e2e. This one will take some time.
These are living notes, not a list of promises.