AI usage generates a fragile codebase

A former Meta engineer argues that AI-generated code often produces brittle, low-quality results that may lead to long-term technical debt in large-scale software development.