whereismyai
Wiggums Proof Loop
Wiggums Proof Loop is an experimental tool that uses AI agents to shrink and simplify Lean proofs repeatedly.
Wiggums Proof Loop is an experiment testing how much AI models can shrink an existing proof in the Lean language. The process involves making changes, checking them, saving the working version, and repeating through multiple rounds. Agents remove repeated code and reuse existing proofs while keeping the protected theorem statement the same. It uses models like Codex and Claude agents working in sequence or parallel with separate working copies. The final proof is checked again and accepted by Lean's kernel using allowed axioms without unfinished proofs.
Wiggums Proof Loop at a glance
Key features
- Repeatedly shrinks Lean proofs using AI agents
- Removes repeated code and reuses existing proofs
- Runs multiple agent rounds in sequence or parallel
- Verifies final proofs with Lean kernel
- Maintains the original protected theorem statement
- Uses allowed axioms without unfinished proofs
Use it for
- Shrinking large Lean proof codebases
- Experimenting with AI agent code reduction
- Optimizing formal mathematics proofs
Alternatives to Wiggums Proof Loop
- Wiggum CLI — Automates development cycles from backlog issues to merged pull requests.
- Aegis — It ensures AI coding agents work against verified baselines and proof.
- CodeGraph — It eliminates file crawling, reducing AI agent tool calls by 88% and cutting costs.
- jevgrep — Reduces coding agent costs by providing targeted, relevant repository context.
- Stagewise — Maintain full sovereignty over AI models while running agents in parallel.
- Pawl — Stops costly agent mistakes using fast deterministic code instead of prompt instructions.
- Loopy — Provides repeatable prompts with checks, proof, and stopping conditions.
- Loop Engineering — Design agent loops to discover work and verify results automatically.
- Axiomcodegraph — Resolves calls from types rather than syntax for higher accuracy.
- Fable Method — Reduces agent hallucinations by enforcing a strict verify-by-observation loop.
- Rigmark — Measures local AI performance with reproducible, agent-like workloads.
- Ponytail — Cuts unneeded code and dependencies by enforcing a minimal-code ladder.
All Wiggums Proof Loop alternatives