Adopting AI Atomically

When you start using AI in agentic mode while programming you quickly realize that you lose grasp on what is actually happening. I've seen people using agentic AI with guides, plans, rules, and the like.. I've tried that and it seems like lawyer work, it takes a long time to process, and I get bored.

A better AI adoption approach could be single line changes with AI. Where you might say "You could've just wrote that yourself". But if you can increase productivity by even 1% by adopting AI then you're setting yourself up to be ready if there's ever a reason to use AI more expansively.

With that said it can be more line changes, that's not the point. The point is the moment the AI makes an unreviewed decision is the moment it becomes unfamiliar. I don't like that as I believe it creates more work than what would be if I made all the decisions myself. So more changes can work if you can review and make decisions on the work rapidly.

Why adopt AI at all?

The problem is about falling behind. I don't necessarily need AI however my choice to adopt it is out of fear and curiosity. Becoming lazy is that last thing I want to do. And the thought of increasing productivity is intriguing.

The challenge is adopting it while also not losing your sense of agency. Perhaps atomically is the way to go.