Developer creates open-source tool that runs multiple AI implementations in parallel to improve code quality through agent-based review and synthesis.

The slot-machine system executes N different code implementations simultaneously, each reviewed blind by separate AI agents, with a judge selecting or combining the best results

Supports multiple AI skills (CE, superpowers like test-driven-development) and harnesses (CC, Codex) with customizable emphasis on different quality dimensions like functionality versus robustness

Leverages the probabilistic nature of AI models—same specs produce different code each time—to increase output quality through parallel execution and intelligent selection

Trades increased computational time and token usage for production code quality, making it cost-effective for long-running overnight processes on subscription plans

Provides a practical method for teams to empirically test which AI skills and harnesses work best for their specific codebases