The pattern I've noticed building tooling for accountants: automation rarely removes jobs, it changes what the job looks like.

The bookkeepers I work with used to spend hours on manual data entry. Now they spend that time on client advisory work. The total workload stayed the same - the composition shifted toward higher-value tasks.

Same dynamic played out with spreadsheets in the 80s. Didn't eliminate accountants - it created new categories of work and raised expectations for what one person could handle.

The interesting question isn't whether developers will be replaced but whether the new tool-augmented developer role will pay less. Early signs suggest it might - if LLMs commoditise the coding part, the premium shifts to understanding problems and systems thinking.

Machine learning is nothing like integer programming. It is an emulation of biological learning, it is designed explicitly to tackle the same problems human minds excel at. It is an organism in direct competition with human beings. Nothing can be more dangerous than downplaying this.