the problem with this argument is that it assumes a linear relation between AI spend and developer productivity.

In reality, it’s log-like. You quickly hit a point of diminishing returns where spending more on AI doesn’t result in increased productivity. Coding might no longer be the bottleneck, but human thinking still is.

SWEs still need to review and own the code. They still need to think about problems and solutions. Those don’t scale linearly with AI coding budgets.