LLMs and performative productivity
Summary
An opinion piece that challenges the notion that AI, specifically LLM agents, universally boosts software development productivity. It synthesizes multiple studies showing mixed results, highlights cognitive debt and maintenance costs, and argues for a holistic, long-term view of productivity beyond speed and PR counts. The author concludes that AI tools can be useful when applied judiciously, but cautions against letting marketing or short-term gains distort the true value of work.