The Dependency Gradient

The Feeling Nobody Names

There's a specific feeling that comes after a heavy AI-assisted coding session.

You're done. The PR is up. The tests pass. The deploy is green.

But underneath the completion there's something smaller, quieter: a question you don't want to ask yourself.

Did I actually build this, or did I mostly manage the AI's output?

Most engineers I know feel this. Almost none of them talk about it.

Not because they're ashamed — though some are. Not because they think they're alone — though they assume they are.

They don't talk about it because naming it feels like it might make it real. Like if you said it out loud, you'd have to do something about it. And doing something about it sounds exhausting.

What the Velocity Signal Misses

Here's the uncomfortable part:

Your shipping velocity is probably going up. The output increased. The signal said "you're doing great."

The signal was right about the output. It was silent about the dependency.

The gap between those two numbers — what you can demonstrate with AI, and what you can do alone — is the thing nobody tracks. Until they suddenly have to.

What Nobody Tells You About Recovery

If you've noticed the gradient shifting — if you've felt the line moving — here's what I want you to know:

You don't have to choose between using AI and maintaining your skills. Those aren't actually opposites.

The engineers who navigate this best aren't the ones who use AI least. They're the ones who've gotten intentional about where the line sits for them specifically — and which skills they're willing to let migrate and which ones they want to keep on the "can do unaided" side.

This requires two things most engineers skip:

• A: Naming the skills that matter to you. Not commercially. Not by what the industry says you should know. What actually matters to you — what you'd miss, what you'd feel bereft without.
• B: Protecting them deliberately. Not by avoiding AI entirely. By carving out specific zones where you don't use it — not because efficiency doesn't matter, but because some things are worth maintaining at depth.

The boundary is personal. The boundary is also learnable.

The Thing Worth Protecting

Your confidence isn't the same as your capability — but they're related. When you maintain the skills, the confidence has a real foundation. When the skills drift without you noticing, the confidence can feel increasingly theoretical.

This isn't about "earning" the right to use AI. It's about knowing where you actually stand.

The clearest engineers I know have made a kind of peace with the gradient: they've accepted that some things they'll maintain at depth, some things they'll use AI for, and they're clear about which is which.

That clarity is underrated. It reduces the ambient anxiety that comes from not knowing whether you could still do the thing — or whether you're just good at directing the AI that does the thing for you.

You can know. The 20 minutes will tell you.