A lot of software engineering career capital was built on knowing which obscure method to call, which Stack Overflow answer to trust, how to navigate a specific framework's quirks. That knowledge was genuinely hard to acquire and it was a real signal. Now it's table stakes. The career capital that survives is knowing why you'd make a particular architectural decision, how to tell if generated code is actually correct, what the error message is really telling you.
The road warrior framing is right. Those people internalized systems thinking across the whole stack over years. AI doesn't replace that — it makes it worth more, because now one person with that mental model can move faster than a team without it. The people who are "bored of AI" are often the people who already made that transition and stopped finding it novel. The people still anxious about it usually haven't yet.