My most valuable and rewarding skill these days is being extremely aware of my current workflow, so whenever I hit a bottleneck I can immediately think of a way to improve it — or at least capture enough context so it’ll be possible to do it later.
Vibe-coding and agent-based processes make the building part infinitely cheaper, but being power-drunk on AI really makes it tempting to try to automate that one bottleneck I think I keep hitting.
I’ve found that I get the real ROI from identifying and solving a real problem I’m facing this very moment, since the justification for doing so — and therefore, the exact success criteria for any new tool — is literally staring me in the face.