Ralph Wiggum loops

A deterministic AI development methodology that uses persistent Bash loops to force LLMs into continuous self-correction until tasks meet defined success criteria.

Ralph Wiggum loops represent a shift from single-shot prompting to autonomous, iterative engineering. Popularized by Geoffrey Huntley, the technique treats an AI agent as a persistent process that refuses to exit until a specific completion promise—such as a passing test suite or a successful build—is satisfied. By feeding the agent's own errors and workspace state back into the next iteration, Ralph loops leverage brute-force persistence to handle complex migrations and greenfield development. This approach effectively removes the human-in-the-loop bottleneck, allowing developers to define architectural constraints upfront and wake up to completed repositories (and occasionally entire programming languages) built overnight.

https://awesomeclaude.ai/plugins/ralph-loop