We are an independent research lab studying recurrent structure and invariants in computational and physical systems, informed in part by Michael Levin's work on cognition across scales, from cells and tissues to artificial and collective systems.
Distributed Flow Lenia search, compendium indexing, and replay tools for studying persistent forms and behavior.
Jolt simulations that compare local material-memory updates against memory-free and damping-only controls.
Sorting-algorithm dynamics analyzed through mechanistic signatures, perturbations, and robustness assays.
Intervention framework for proof-search structure: centralized MCTS maps proof families, basin stability, and search efficiency, while distributed MCTS tests how collective control changes access to that landscape.
Benchmark harness for Lean sorry tactic proposals, with split policy, scoring, and verification protocols.
Deterministic ingestion pipeline for SorryDB-compatible snapshots and benchmark indexes.
Implementation for reworking CATWD 2.0 semantics and calibrating K across domains to better understand its properties.
Offline training and reranking over exported Wonton Soup run traces.
Compact decision procedure for equational implication over magmas, distilled from the 4694-law implication graph into a sub-10KB offline cheatsheet for SAIR's Stage 1 benchmark.
Intervention studies in proof search, with structural comparison across runs.