Hey!

Chris here. Welcome to Blueprint—the newsletter to help you build a winning engineering team.

I recently had a call with a super sharp guy who’d clearly done his homework. He showed me dashboards he'd been building, dropped the right product names, and knew his way around Claude and the new tooling.

He was also about 6 months into the most predictable arc I've watched play out in the world of AI.

Let me break it down. 👇️ 

📒 DEEP DIVE

The Euphoria Trap

How AI makes building feel effortless—until performance, security, and maintenance show you what’s missing.

Like Lightning in the Fingertips

There's a path people follow when they start using these tools. I've been watching it long enough now that I can almost set my watch to it.

It starts with what I'd call the technical-but-not-engineer crowd. Founders, CEOs, product owners, project managers—people curious enough that AI piques their interest rather than immediately shutting their brain off.

When they get their hands on something like Claude—which has a fantastic personality, describes things exceptionally well, and is extraordinarily confident—something clicks.

They hit euphoria.

They start building with these tools, and they feel like they have lightning flying out of their fingertips.

And honestly? A lot of that feeling is justified. The barrier between idea and working prototype has collapsed in ways that weren't possible 2 years ago, and that's exciting.

The Wall

The problem is all the things they don't know don't show up at the beginning. It's not until the first time they release something to real users and the site isn't performing the way they expected or there's a security issue.

Pretty soon, they're spending every single day manually prompting Claude over and over to handle support problems—which puts them right back in the exact seat they were trying to get out of.

They wanted to build and to create. They did not want to become a software support engineer.

And this isn't something AI can solve, because AI hasn't done a single thing to human nature or to human attention capacity.

What you can move through the pipe in a given time frame has gotten bigger, sure. But you still only pay attention to one thing at a time.

And that's the moment when the euphoria starts wearing off.

Too Much of a Good Thing

I read part of an article a while back—at this point it's so blended with my own thinking that I can't tell you where the article ends and my own ideas begin—but it put the right frame on something I'd already been watching in real time.

Think about what it would feel like to live through an era where food went from chronically scarce to suddenly available in enormous abundance.

At first, it's incredible. Everything looks great, and you eat accordingly.

But if you gorge yourself for the next 5 years straight, there are health consequences. There's a calorie budget at work, whether you're paying attention to it or not.

AI has handed people some of what they always thought they wanted, in genuinely massive quantities. You can think of a piece of software and actually build it. That's incredible.

But it doesn't mean it's a good piece of software that should've been built. And it certainly doesn't mean you have enough attention left over to take it from prototype to something you can deploy securely, maintain over time, and add features to without it falling apart.

We are collectively gorging right now. And I say that with full understanding of why—I've been in the same place once or twice, and this is my chosen profession.

The difference is that after 20-something years in this space, I know how to catch these tools at their lies. I know what overconfidence looks like when I'm staring at it.

The Overconfidence Trap

Claude is probably the most confident software I've ever worked with.

It is genuinely exceptional at UI design—I'd go so far as to say it's currently the best available for that. It's also a tremendous teacher and very good at explaining architecture.

Deploying architecture and back-end engineering in general are areas where it consistently falls short.

When someone spends 6 months fully inside that confidence—treating it as gospel and building on top of it without the background to know what to question—there's a simple way to stress test where things actually stand:

Take a completely separate model, one that has never seen what was built, and have it do a code review.

What it surfaces is usually not pleasant, and there's a reason for that.

If you ask the same agent that built the thing to critique it, the results are awful. It's going to protect its own work the same way any engineer would. You need a model that's never seen the project to do the review.

Most people haven't learned to do this yet. And that's fine. We're early.

Humanity is still working out where the real edges of this thing are, and that process takes time. There's going to be a lot of good that comes out of it on the other side.

BEFORE YOU GO…

If you haven't hit the wall yet, you will. I don't mean to sound threatening—that's just what the arc looks like.

Euphoria first. Then the gaps show up. Then you figure out what you actually have.

Everyone goes through it. The ones who come out the other side are just the ones who kept going.

Talk soon,

Chris.