December 6, 2025
Infinite Loops, Finite Patience
Running Claude Code in a loop to mirror human development practices
Endless AI coding sprint sparks cheers, side‑eye, and a Ralph Wiggum meme
TLDR: A dev made Continuous Claude, an AI that keeps coding in a loop and remembers progress to push test coverage higher. Commenters are split between admiring the clever relay approach and worrying about client consent and test quality, with memes crowning it the Ralph Wiggum of coding—persistent and chaotic.
A developer, drowning in unit tests, unleashed Continuous Claude—a loop that keeps Claude coding, opens GitHub pull requests, and leaves cliff notes in a markdown file so the AI remembers what happened last time. Think relay race, not sprint, with each pass making small progress and notes for the next lap. The crowd loved the audacity but side‑eyed the infinite "while true" loop.
Strongest reactions hit fast: yellow_lead demanded, “Is your client ok with this? Are the tests any good?”—the ethics-and-quality alarm. Others dropped receipts, linking rival loops like claude-loop and the official Ralph Wiggum plugin, named after the cartoon kid who keeps trying despite setbacks.
The jokes wrote themselves. Folks dubbed it CI/CD (automated build-and-deploy) for vibes, and memed Claude as Ralph yelling “I’m helping!” while failed pull requests get closed and tossed. Fans adore the single-file memory trick—one markdown to stop the AI from forgetting—calling it genius and hilarious. Skeptics say it’s wasteful and risks brittle tests. Meanwhile, GitHub Next’s "Continuous AI" got shout‑outs, signaling that loop‑life might be the next big trend in coding help.
Key Points
- •Continuous Claude is a CLI tool that runs Claude Code in a loop with persistent context to improve large codebases, such as increasing test coverage.
- •The system integrates with GitHub pull request workflows, respects CI checks and approvals, and can be scheduled or triggered.
- •A Bash script orchestrates iterations, using GitHub CLI (gh pr checks); failed iterations close PRs to reset and try alternate approaches.
- •Context is persisted via a shared markdown file, with model instructions to make incremental progress and leave clear handoff notes, reducing context drift.
- •The approach aligns with continuous AI development explored by GitHub Next’s Continuous AI, and the tool was shared with that team.