Back to HomeResearch Computing Platform
Collatz Lab
Local-first Collatz research workspace with CPU/GPU runs, validation, claims, source review, reports, and live math visualization.
10 technologies
12 features
About this project
Collatz Lab is a local-first research platform for the Collatz conjecture. It is explicitly not a proof claim - it is a workspace for reproducible CPU/GPU experiments, validation, claims, source reviews and evidence tracking.
The platform provides backend orchestration, SQLite persistence, a FastAPI API, CLI, resumable CPU/GPU worker loop, compute profile control, and a React/Vite dashboard with overview, evidence, operations, live math and paper views.
The project is designed around correctness boundaries: validated computation is evidence, not proof. AI is optional and limited to review/planning assistance.
Key features
Compute & workers
- CPU/GPU experiment runs with queueing and replay
- Resumable workers with checkpointed execution
- Compute budget controls (system/CPU/GPU percentages)
- Hardware-aware execution: CPU, CUDA GPU, Apple MPS, native Metal
Evidence & claims
- Run validation pipeline
- Claims and claim-run links
- Source review with fallacy tags and consensus baselines
- Artifacts and lab report generation
Dashboard & research
- Dashboard tabs: overview, evidence, operations, live math, paper view
- Hypothesis sandbox: battery tests, residue class analysis, record seeds, trajectory depths
- Native/GPU benchmark diagnostics with presets and hall-of-fame
- Optional Gemini/LLM review and planning assistance
Collatz Lab research workspace dashboard