Métricas de qualidade de software na era da IA

This is an editorial piece, not a tool, but it addresses something most engineering teams using AI code generation have not formally sat down to think through: the quality metrics designed for human-authored code break in subtle ways when a significant portion of your codebase is model-generated. The article argues that coverage numbers can look healthy while the actual logic is brittle, and that cyclomatic complexity loses meaning when a model can generate deeply nested code that passes all tests and still behaves unexpectedly at the edges. The framing is practical rather than academic — what should a team actually measure and how. Reservation: it is in Portuguese, which limits the audience here, and it is an article rather than an actionable framework or tool. The ideas are worth the translation effort if your team is already shipping AI-generated code to production and has not revisited your quality gates. -> Best for: technical PM or SaaS team of 2-5 with AI-assisted development already in their workflow