The author argues that AI friendliness should be prioritized in software projects, proposing that tools like API mocking servers be designed with AI as a primary consumer rather than just human developers. They cite their project Apate as an example, suggesting AI should handle writing mock API specifications while developers focus on main application logic.
Meta's experiment with smaller AI-supported engineering "pods" of three to five people increased code changes 220% but only raised shipped features 36% while incidents jumped 40%. Research across 500+ organizations found similar patterns: pull request throughput rose 37% while developer experience fell and engineers spent more time firefighting problems. Optimal team size remains five to eight people, with shrinking best done gradually after identifying which tasks AI can safely handle.