For me, one signal has been whether the problems remain interesting even when progress is slow.

When working on complex systems (like anything involving long-running automation or agents), most of the real work happens in areas that don’t show up in demos: defining “done”, handling partial failures, and keeping behavior predictable.

If those problems are still worth thinking about after repeated failures, I take that as a sign the work itself is worth continuing.

That resonates a lot.

I’ve noticed the same thing: if the shape of the problem is still interesting after the novelty wears off and progress stalls, that’s usually the real signal. The visible demo work is easy; the hard part is exactly what you said—defining “done,” handling edge cases, and making systems behave consistently under stress. If those invisible constraints keep pulling you back, it’s usually because there’s something fundamentally worth building there.