Wonton Soup Article
Long-form overview of the project, methods, and results to date.
Wonton Soup
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.
Research Questions
When we lesion a solver's search process, blocking a tactic or altering coordination, does it return to the same proof structure or settle into a different one? Do proof spaces exhibit multistability under fixed configuration and controlled perturbation?
Can search efficiency (K) be measured on a shared scale across substrates, following the Chis-Ciure and Levin framework for comparing directed search to blind baselines? Does collective control change which regions of the landscape are accessible, and does collective search create efficiency that individual search does not access?
Start Here
- Read the dossier framing article.
- Use the Wonton Soup docs map for concepts, schemas, and runbooks.
- Open the dashboard for run-level and theorem-level inspection.
Latest Updates
- March 1, 2026Cabinet docs rebuilt with updated Wonton Soup references and stable doc IDs.
- February 6, 2026Dashboard manifest defaulted to run
corpus-2026-02-06-003808with 522 theorems. - February 3-12, 2026Follow-up analysis window logged with 355 run entries and intervention taxonomy pressure tests.
Key Outputs
Documentation
Docs, ADRs, and runbooks for the dossier.
Dashboard
Interactive dashboard for run comparison and metric inspection.
Project Source
Full codebase, CLI entry points, and test harnesses.
Data & Artifacts
Dashboard Manifests
- Dashboard manifest (run selection and metadata).
- Active payloads live under the dashboard data directory and are selected by the manifest rather than a hardcoded run id.
Cabinet Contracts
Dashboard