[Nathanf22]

One angle that's often missing from these articles: what happens to the codebase after AI writes the code. The articles I've seen focus on velocity — how fast AI generates features. But the harder conversation with managers is about comprehension: AI can produce 10x more code than a human, but nobody has built the tooling to understand the 10x more complex system it creates. The best framing I've found for non-technical managers: "We're getting better at writing code. We're not getting better at reading it." That gap is where the real risk lives.