When AI misses: take over or keep pushing?

Some of the biggest resistances to AI aren't technical. To my mind, they're psychological.

A few weeks ago, I gave a task to one of our experienced devs. Automation stuff — take content from one channel, translate it, repost to another. If someone puts a checkmark in a Google Sheet, push it to a third channel. Not rocket science, but not trivial.

This guy has 10+ years in the field. He uses AI. But I noticed something watching him work. He'd start with Claude, get some code back — and then rewrite it manually. This happens constantly. Looking at his commits, it seems like he wants AI to write code the way he writes it.

Around the same time, I gave a similar task to my son Zhenya. Still in university, good technical school. Comparable complexity.

Completely different approach. He described the task to Claude and started iterating — using what they taught him at university. How to prompt, how to verify. He barely looks at the editor. Just keeps pushing until it works. Pure vibe-coding. 🤷

I'm sure he lacks the experience to have a picture of what "proper structure" should be. So he just doesn't know any other way than to push forward. When AI writes something unexpected, it looks more like "not done yet" than "this is wrong."

Code quality? Hard to compare. Both had room for improvement. But both got to a working result. Though Zhenya got there faster — not dramatically, but noticeably.

The difference isn't about who's smarter. It's what happens when AI doesn't deliver. One takes over. The other keeps pushing.

What does this mean for business? Maybe a junior with the right education costs less and moves faster. The question keeps bothering me — what exactly are we paying for when we pay for "experience"? 🤔

Is there risk in 100% vibe-coding? Yes. Does it work for all tasks? No. But for some — it already does.

This isn't a conclusion. It's an observation. 👀

What about you? When AI doesn't give you what you expected — do you take over or keep pushing?