Let’s talk about LLMs

A detailed critical analysis of LLM coding tools through the lens of Fred Brooks' 'No Silver Bullet' essay. The author argues that LLMs primarily address 'accidental' difficulty in software development, not the 'essential' difficulty (specification, design, testing of conceptual constructs), and therefore cannot deliver order-of-magnitude productivity improvements. Drawing on the DORA report, CircleCI's 2026 State of Software Delivery, and the METR study, the post highlights that empirical data shows increased delivery instability, higher failure rates, and self-reported productivity gains that don't match reality. The author also challenges the 'democratization' narrative, arguing that effective LLM use still requires deep programming knowledge. The conclusion recommends focusing on software development fundamentals (version control, CI, testing, small batches) rather than rushing to adopt LLM tools out of fear of being left behind.