What problem does it solve?
Iterative convergence of AI-generated code by running adaptive denoising loops that transform chaotic outputs into a coherent product, stopping only when entropy reaches zero.
Core Features & Use Cases
- Iterative denoising loops: measure residual entropy, target gaps, implement fixes, and re-denoise to converge to a complete solution with bounded iteration scopes.
- Convergence governance: supports supervisor/watchdog mode and specialized agent-based execution patterns (vc-delegate and vc-agents) with structured loop reports and convergence metrics.
- Robust state and safety: frontmatter-driven iteration limits, optional completion promises, and per-session state tracking to ensure repeatable progress across runs.
Use cases include iterative refactoring, incremental feature implementation, and systematic bug fixes in AI-assisted software projects.
Quick Start
Start a Marbles loop in your codebase and let it iteratively refine the output until convergence is achieved.