One thing I've noticed while working with AI tools is that they struggle with many of the same things junior engineers struggle with.

Not algorithms, and syntax, nor frameworks.

They struggle with ambiguity.

Give an AI vague requirements and you'll get vague results.

Point it at a codebase with inconsistent abstractions and it will make incorrect assumptions.

Ask it to work with undocumented business rules and it will confidently fill the gaps with guesses.

For years, teams survived these problems because people carried the missing context in their heads. New engineers asked questions. Senior engineers explained the history behind certain decisions. Knowledge spread through conversations.

AI doesn't have access to those conversations.

It only sees what's actually written down.

That's why, looking at some of the basic requirements for working with ai tools, I think AI is doing something unexpected: it's making software quality more visible.

A clean architecture isn't just easier for developers to navigate. It's easier for AI tools to understand.

Clear requirements aren't just good project management. They're becoming a prerequisite for effective AI collaboration.

The interesting lesson isn't that AI is imperfect.

It's that many of the things confusing AI were already confusing humans.

We just got used to working around them.