THE 80% PROBLEM IN AGENTIC CODING

Summary from the TLDR newsletter

AI writing most of the code places a disproportionate emphasis on the human's role of owning outcomes, maintaining quality bars, and ensuring that tests actually validate behavior. The danger is that the AI succeeds so well that developers stop checking their outputs. AI is an amplifier of development practices. Bad processes will result in accumulated debt at an unprecedented speed.