Project
See all projects on GitHub →
Discovery engines for proofs, programs, and patterns.
We build reflective systems that learn from their own reasoning—spanning program synthesis, metaprogramming, and neurosymbolic search.
What we build
Reflective tooling
Systems that inspect and steer their own search—proof by reflection, meta‑interpreters, and verified tactics.
Synthesis & search
Enumerative and neural‑guided synthesis, constraint solving, and structure discovery for code and theorems.
Neurosymbolic bridges
LLM + logic hybrids for data‑efficient reasoning, guided refinement, and interpretable proof‑plans.
Open research artifacts
Readable code, docs, and demos—made to be forked, remixed, and falsified.