Stop Learning AI. Start Using It.

Every engineer I know has the same confession: "I feel like I'm falling behind on AI."

Senior engineers. Mid-career engineers. People who've been writing code for a decade. All of them feel like they missed something. Like there's a course they didn't take, a concept they haven't internalized, a skill gap growing wider every week.

Here's what I've figured out: that feeling is a trap.

The engineers who are good at AI didn't study their way there. They just started using it — for everything — until it became instinct. The gap isn't knowledge. It's habit.

The Mindset That's Holding You Back

Most engineers approach AI like they approach a new programming language. They want to understand it first. Read the docs. Take a course. Get comfortable before they commit.

That's the wrong mental model.

AI tools aren't something you learn and then use. They're something you use until you've learned them. The reps come first. The understanding follows.

The Pragmatic Engineer recently noted that engineering leaders are talking about this behind closed doors: mid-career engineers are feeling the most left behind — even more than junior devs. New grads just started using the tools without overthinking it. That's the whole edge.

The Only Rule That Matters

AI first. Every time. No exceptions.

Whatever you're about to do — copy a file, write a regex, remember a terminal command, debug an error, Google something — stop. Ask AI first. Even for things you already know.

Especially for things you already know.

That sounds extreme. It is. That's the point. You're not optimizing for efficiency right now. You're building a reflex. The reflex is the thing.

What This Actually Looks Like

Here's what "AI first" looks like in a real day:

Terminal commands → Before you try to remember the rsync flags, ask Claude. Every time.

Debugging → Paste the error directly into your AI tool before you Google it. Don't read the stack trace first — let AI read it with you.

New codebase → Ask for a tour. "What does this repo do, how is it structured, where should I look first?" Don't grep blindly.

PR descriptions → Draft in AI, edit yourself. Never start from blank.

Feature planning → Spend time in "plan mode" before you write a line of code. Talk through the feature, the edge cases, the architecture. Think of it like rubber duck debugging — but the duck talks back.

Anything you'd normally Google → Ask Claude instead. The answer will be better. More importantly, you'll start routing your thinking differently.

None of this is complicated. That's the point. The barrier to entry is zero. The only thing stopping you is the habit of not doing it.

Here's What Happens When You Actually Do This

Last week I shipped four apps and websites. One night each. Most of the code was written by AI while I was asleep.

I'm a software engineer at Meta. I know exactly how long this used to take.

The shift wasn't that I suddenly understood how LLMs work. It wasn't a course or a certification. It was that I'd already built the muscle. I'd been asking AI for everything — small things, obvious things, things I didn't need help with — until the instinct was automatic.

When it came time to build something real, I didn't have to think about the tool. I just used it.

That's the whole game.

Where to Start Tomorrow

Not a course. Not a YouTube video. One habit.

Next time you're about to Google something — anything — open Claude instead and ask it there.

That's it. Start there. Do it ten times tomorrow. Do it a hundred times this week.

The engineers who feel behind aren't missing knowledge. They're missing reps. Go get the reps.