I first heard this SJ quote maybe 15 years ago and I've carried it with me since.

I've always prided myself on taking a craftsman-like approach to software engineering... thinking deeply about interfaces, ownership, lifetimes, how the public API looks, how using the public API feels...

Lately, though, with the advent of LLM-assisted coding this mindset is starting to feel hollow. Why spend 1.5x as long crafting something robust when, in all likelihood, it will be replaced or refactored by LLM tooling within the next 5 or 10 years?

Unless LLMs significantly improve, LLMs benefit greatly from stable APIs. Constantly changing APIs are awful for good LLM output.

is this the right way to think about LLMs?

For example, the syscalls for linux are never changing IMO. The cost is unbounded to change, even with AI.

Should your APIs be treated any different?

At the very least, the LLMs work better with better APIs and data models, which yet accelerates the solving of problems.