Smoothing the Path with AI

My favorite use of AI at the moment is not vibe coding toy applications but setting up systems. A test suite that would have previously taken 2 weeks to configure end-to-end now only takes a few hours while I’m doing something else in the background. I’ve found the biggest time saving to be in exploring radically different approaches, searching through piles of settings, and resolving various compatibility issues.

I could spend hours digging through documentation, StackOverflow, or forums before I finally unearth something useful. Now AI can search everything for me, find a solution, try it out, then try a different one if that doesn’t work out, and finally come back with the issue fixed.

On the surface there seems to be less learning now, but did this process really lead to that much learning to begin with? Were most of these hours of research not spent on grappling with useless leaky abstractions, and then once I finally solved the problem - was I not relieved to immediately forget about it?

Perhaps there is something to lose. I did eventually learn most of the gotchas and know what to steer away from, though this knowledge was never really all that useful for achieving my goal of building reliable software. Yes, AI can smooth the path to the actual task, like driving a car on a highway versus cross-country hiking. Yet when we go hiking we’re more adaptable to obstacles, we exercise, we learn the way of the land - all of these have value on their own, and there is still purpose in pursuing them. We should still occasionally try to make difficulties everywhere.

So every day we have a choice to make - what’s more important right now: to get to know the land, or to get from A to B? It is for each of us to decide. But if we choose the smooth path instead of the hike every day, we may be losing something elusive and difficult to measure.