Disclaimer: I hate writing. I’m using AI to get my ideas onto paper. The opinions, experience, and numbers are mine. The grammar is not.

Musings of an overtired CTO dictating into his phone with a sleeping baby in his arms.

I keep seeing the same post on LinkedIn. Some variation of: “I don’t code anymore. I just review AI output. My skills are atrophying. The worst part is AI is coding better than I did.”

Three sentences. A career, a craft, and an identity, all renegotiated in real time. I get it. But I think the response to that feeling is completely wrong.

I’m Writing Rust Libraries From My Phone

Right now, at this moment, I am writing code from my phone with a newborn in my arms. I’m experimenting with performance tuning GIF encoders. I’m learning about SIMD instructions. I’m exploring different quantization algorithms. I’m wrapping Cloudflare’s Pingora into a full L7 proxy engine to compete with Envoy. I’m building eBPF and XDP toolchains. I’m creating things that would have taken me weeks or months, and I’m doing it in days. From my phone.

This is not a flex. This is the point.

We are living in one of the most innovative periods in the last twenty years of software engineering, and most people are spending it feeling sorry for themselves. You have a pocket-sized, hyper-focused, relentlessly productive junior developer available to you twenty-four hours a day. It doesn’t sleep. It doesn’t get bored, but it’ll sometimes get lazy. It costs you two hundred dollars a month. And it will try any idea you throw at it, no matter how insane.

You want to challenge an assumption? Challenge it. You want to ask “what’s the nuclear option for solving this problem?” Ask it. The autistic developer in your pocket will go try it. Right now. While you’re on the toilet.

The Secret Is in the Tests

I was never a big fan of testing. I relied on my own skill to catch problems. That worked when I was the one writing every line. With AI, it doesn’t work. You have to test. You have to provide guardrails. You have to force the model to think.

But here’s the subtle thing. When you force AI to write tests, you force it to write code that isn’t entangled. Testable code can’t be spaghetti. Encapsulation, separation of concerns, clean interfaces: these aren’t just best practices you lecture junior developers about. They’re the natural byproduct of making code testable. Testing is the constraint that turns AI slop into real engineering.

Not making AI slop is the same skill as not writing spaghetti code as a junior developer. It’s knowing how to ask for encapsulation. It’s knowing how to demand separation. It’s knowing how the pieces should fit before you let the model touch them.

Pick One

The mentality of “I’m just a code reviewer” or “I’m just a slop engineer” is defeatist. It’s also a choice. You have the tools at your fingertips to compete with the best in the world. You have the ability to experiment, to learn, to build things that didn’t exist yesterday. You can pick up computer science fundamentals you skipped. You can study the people who wrote the books on code design. You can prove what good output actually looks like.

Yeah, it’s better to do it from a computer. But if you’re busy, if you can’t spend eight hours at your desk on top of your day job, you can still build novel things and challenge assumptions and learn. The barrier to entry for ambitious engineering has never been lower. The variable isn’t years of experience. It’s curiosity and craft.

The people who see AI as a threat to their identity are going to get passed by the people who see it as the most powerful learning tool they’ve ever held.

Pick one.

You’re not a review bot. You have a mass-produced savant in your pocket that will try any idea you throw at it. Use it or watch someone else do it.