Your Engineering Career Was Never as Safe as You Thought

Dan had a therapy session with Claude this week.

That’s not a joke, and it’s not a setup for one. He sat down, opened a conversation, and worked through the thing that’s been sitting in the back of a lot of engineers’ minds lately: what happens if AI just does everything I used to do? For twenty-plus years his career had been a solid foundation. He could walk into almost any company and get a job whenever he wanted, wherever he wanted. He had control. And lately that ground has felt like it’s crumbling underneath his feet.

What Claude told him is the part worth sitting with. It wasn’t reassurance. It was that the security he’d felt for two decades was a false sense of security the whole time. His career wasn’t actually as stable as he believed — it just hadn’t been tested yet. The disruption didn’t remove his safety. It revealed that the safety was never as guaranteed as it felt.

That reframe is the most useful thing we’ve heard about AI and careers in months. So let’s take it seriously.

The story your brain is still telling you

Here’s the trap. If you’ve been good at something for a long time, your brain encodes “I am safe” as a permanent fact about the world. Dan was, by multiple people’s accounts, the best DevOps engineer in the room for 25 years. That’s not bragging — it was just true. And after 25 years of it being true, his nervous system stopped treating it as a current assessment and started treating it as a law of nature.

Then the ground moved. And the hardest part isn’t the new reality — it’s that his mind has been trained for two decades to expect the old one. He’ll tell you himself: “I literally am not a DevOps engineer anymore. I don’t even know what I call myself.” The skills didn’t evaporate. The identity did. And identity is stickier than skills, which is why this transition messes with people far more than learning a new framework ever did.

This is the quiet thing nobody puts on a slide. The anxiety engineers feel right now isn’t really about whether they can learn the tools. Of course they can — they’ve learned harder things. It’s about grieving a version of themselves that felt permanent and turned out to be contingent.

”Am I gonna be a waiter in a year?”

We say this with full awareness of how it sounds: we are about as all-in on AI as two people can be. We restructured our entire working lives around it. And even we have the intrusive thought — okay, am I going to be a waiter a year from now?

We’re not telling you that to be dramatic. We’re telling you because if the people who are most bought in still feel the fear, then the fear is not a signal that you’re behind. It’s a signal that nobody can see the future, including the optimists. The honest answer to “what happens in five years” is we don’t know. Anyone selling you certainty in either direction — total doom or total safety — is selling.

What we can see is the shape of the present. And the present is moving.

Why “I’ll learn it later” is a worse bet than it sounds

The most common rationalization we hear is reasonable on its face: “If it’s changing every three months, why learn it now? In three months it’ll be completely different and I’ll just learn it then.”

Maybe. We genuinely can’t rule it out. But here’s the asymmetry. The cost of engaging now is twenty bucks a month and some evenings. The cost of waiting is that the gap compounds while you wait — not the tool knowledge, which does churn, but the fluency. The intuition for how to plan work for an agent, where it’s strong, where it lies to you, how to structure a repo so it can actually move. That doesn’t reset every three months. It accumulates. The people who started a year ago aren’t ahead because they memorized an interface that no longer exists. They’re ahead because they’ve built judgment that transfers across every version.

So the bet isn’t “learn the current tool.” The bet is “start building the judgment that survives the tool changing.” Those are very different things, and only one of them gets cheaper by waiting.

The line that’s going to make people mad

Here’s the sharp version, and we’ll own it: if you’re choosing to not use Claude Code right now, you are self-selecting out of a career.

Not because a manager will fire you for it. Because the field is reorganizing around a new baseline of what one engineer can produce, and opting out of that baseline is a choice with consequences — especially if your company has already embraced it. And companies are embracing it across the spectrum we used to think of as the safe harbors. Government work. Payments. Healthcare. The compliance-heavy corners where the conventional wisdom was “this will never change, you can hide here for a decade.” It’s changing there too.

We want to be careful here, because there’s a lazy version of this argument that blames AI for every layoff, and that version is wrong. The big cuts at the large tech companies? We don’t think those were mainly about AI. We think a lot of them are balance-sheet decisions — post-pandemic tightening, a market that rewards profit over growth, companies that simply over-hired. Blaming AI is often a convenient story. Notice who isn’t doing layoffs: the companies going all-in on AI are hiring. That’s the tell. The narrative “AI is taking the jobs” and the data “the AI-forward companies are growing headcount” don’t fit together cleanly, and you should be suspicious of anyone who insists they do.

There’s a real fear underneath it, though, and it deserves a straight answer. Engineers hear that a company adopted AI the right way, started moving fast, and now has a half-million-dollar-a-month model bill. And the immediate thought is: they’re going to offset that by laying people off. It’s a fair worry. But run it forward. If that half million buys you moving twice as fast as your competitor, you don’t cut the people producing the speed — you press the advantage. The companies that treat AI as a cost to claw back will behave very differently from the ones that treat it as leverage to extend. The first kind might cut. The second kind is the one hiring. Choose your employer accordingly.

What we’d actually tell an engineer today

If you came to us — junior, senior, doesn’t matter — here’s the whole of it.

Embrace change, and go all in. Not as a motivational poster. As the literal best-EV move when you can’t predict the outcome. We got into tech because things change and that’s what makes it interesting; this is just the most extreme version of the thing that always drew us here. Dan will tell you it reinvigorated him — there’s so much to learn right now that the job got fun again.

If your company hasn’t embraced it, route around them. Spend the twenty dollars a month yourself. Build on your own time. Or — and we mean this — go find a company that is embracing it and go there. Get ahead of the curve while the curve is still gentle.

And build the dumb stuff. This is the part people skip, and it’s the part that actually rebuilds your confidence. How many side projects have you wanted to build and never had the hours for? Go ask Claude to build one and watch what happens. Dave has seven Solitaire sites, a fly-fishing site, and a fan site for his brother’s novels — one of which he built on a ski lift in Mammoth, in a blizzard, between runs with his daughter. By the end of the day there was a working website. Was it important? No. Did it rebuild the felt sense that he can make things again, fast, without the old friction — the pull requests, the commit messages, the README upkeep that used to sit between him and the actual fun? Completely.

That last point is the antidote to the identity grief. You don’t argue your way out of feeling obsolete. You build your way out of it. The fear says “the thing I was good at is gone.” Shipping something — anything — in an afternoon answers back: “and the thing I’m becoming good at is faster and more fun than the old one ever was.”

The honest close

We can’t read the future. We’ve said it three times now because it’s the only intellectually honest thing to say, and because every confident prediction you’ve heard this year has an expiration date. Maybe the optimists are right. Maybe the worriers are. Probably it’s some messy middle nobody’s describing yet.

But Dan’s therapy session landed on something that’s true regardless of which future arrives: the security you felt was always more contingent than it appeared. AI didn’t take a stable thing and make it unstable. It took an unstable thing and showed you it was unstable. That’s actually good news, because it means the move now is the same move it always quietly was — keep learning, stay close to where the field is going, and don’t mistake “I’ve been safe for a long time” for “I will be safe forever.”

The disruption is here. It’s not stopping. We recommend you go all in — not because we’re certain it pays off, but because it’s the most alive, most interesting, and frankly most fun version of the bet. And because the alternative is hoping the ground stops moving. It won’t.

This is the same lesson we keep running into from the org side, too: the teams that thrive don’t wait for certainty before they start. They start small, build momentum, and let it compound — which is exactly what we wrote about in why you can’t transform 200 engineers at once. The individual version of that truth is just: start now, start small, and let your own momentum compound.

---

This post is adapted from The Velocity Lab podcast, Episode 22: I Had a Therapy Session with Claude.