Vibe Coding Cleanup as a Service (donado.co)

A new niche — “Vibe Coding cleanup” — is emerging as companies pay specialists to fix production-unready AI-generated code. Coined by Andrej Karpathy, vibe coding (developers prompting AI to generate functions) boosts productivity and is now widespread—GitHub says 92% of developers use AI tools—but studies show real costs: GitClear found AI-assisted work correlates with 41% more short-term code churn, Stanford researchers report less secure output despite developers’ confidence, and Computer Weekly flags ~40% of AI-written code containing vulnerabilities. Common failure modes are local, context-free code that introduces inconsistent interfaces, duplicated logic, leaked secrets, deprecated dependencies and race conditions that only surface at scale. That failure pattern has spawned a market: consultancies (Ulam Labs, Donado Labs) and marketplaces (VibeCodeFixers.com) match cleanup specialists to projects, with rates typically $200–$400/hour or fixed-price “vibe-to-production” packages and audits. Gartner predicts 75% of enterprise engineers will use AI assistants by 2028, implying huge demand for remediation. Practically, teams must treat AI as a prototyping accelerant and bake in human-led architecture, testing, and security hardening. The result: new career paths for cleanup experts and a competitive advantage for orgs that standardize remediation pipelines rather than outsourcing long-term maintainability to AI alone.