Certified Shovelware (justin.searls.co)

Developer Justin Searls launched a "Certified Shovelware" GitHub badge and declaration to spotlight projects whose code was predominantly written by AI coding agents, yet still "actually work" and likely wouldn't exist without that help. He defines shovelware with three yes/no questions: does the project fulfill its purpose, was the code mainly produced by an AI agent, and would the project not exist otherwise due to time or expertise constraints. The initiative includes a ready-made Markdown badge ([](https://justin.searls.co/shovelware/)) authors can paste into READMEs to openly disclose AI-driven development. For the AI/ML community this is significant because it reframes the productivity debate from theory to artifacts: functioning, useful software built with code-generation agents. The badge promotes transparency and forces skeptics to confront real-world outputs while acknowledging valid concerns about training-data provenance, labor disruption, and the devaluation of craftsmanship. Practically, widespread labeling could accelerate research into agent-assisted development—measuring maintainability, security, licensing risks, and long-term costs—and shape norms around provenance and tool accountability as AI-assisted coding moves from novelty to commonplace.