I vibe-coded a thing

Intro

I built a working web tool over a weekend without writing code, and it made me feel uneasy.

The problem was pretty simple: I’ve accumulated RSS feeds over the years and wanted to try a new reader. When I started thinking about exporting my collection, it felt like a good moment to clean things up and remove inactive feeds. I couldn’t find a tool online, and the use case felt isolated enough to flex my almost non-existent agentic engineering muscle.

This was the first project I “built” without writing a single line of code. Most of the time, I just talked to agents. I had Gemini write the specs, Claude double-check and refine them, and Codex build it.

You can check it out here: https://www.topiaryopml.com

The part that surprised me

I’d heard about the addictive nature of agentic engineering. How it’s finally time for engineers to let out their inner EM and tell someone what to do, almost exactly the way they want it. They wouldn’t even have to be nice (though you should).

To be clear, I have very little knowledge of web development. There are thousands of people out there who could have done this before me. And if you’re one of them, I can completely relate to how that might feel.

When I hear people online say they built an iOS app in 15 minutes, that makes me uneasy too. I can produce the same outcome, but I almost certainly can’t do it in 15 minutes. But here is the key: I used to get paid for however long it would take me.

The emotional hangover

There’s really no arguing this away: I used to get paid for something that can now be done by an agent much faster. That is scary, plain and simple.

I want to stop short of saying a machine will deliver the same quality. I can’t speak for other disciplines, but even with MCPs and skills enabled, I still don’t think agents are at the level of experienced iOS engineers.

But that might just be a matter of time.

The pain is real. The anxiety is real.

Agents are replacing humans in coding. Unlike agents, humans have bills to pay, families to take care of, mortgage rates to keep up with. This is very real risk. At least the country I live in doesn’t have the greatest reputation for social security or for helping people in extreme situations.

Everything people fought for over years, everything they worked on to become good at and eventually become experts in, can now be accessed by anyone. Is that good or bad? I don’t think we know yet.

The general advice is to focus on the things you can change, rather than the things that are out of your control. Much easier said than done when the whole foundation you stand on feels like it’s being erased in a matter of months.

Immediate Next Steps

Speaking of things you can control: learn the tools. Understand how agents work. Maybe even build one yourself.

Software development was never perfect. There were always tradeoffs. Depending on the size of the team and the experience of the engineers, those tradeoffs can have huge effects on the foundation of the software and, by extension, the quality of the products we build.

Things like edge case testing and performance work, in my 15+ years of experience, always somehow fell short to some degree.

My hope is that the time we gain by not spending the largest chunk of our time in code will be used to improve that foundation. We will likely ship faster, but we will also ship better and more stable.

I also hope this leads to more time to invest in ourselves and in our expertise. I hope agents help us level up instead of making us redundant.

Conclusion

Another unknown, and potentially underrated, part of the equation is what it will do to our wellbeing as engineers if we spend less time programming and more time supervising. Agentic engineering feels addictive, but to what extend is this just novelty and fascination. Programming is not only a means to an end. It is an expression of creativity, and it’s intellectually challenging.

I’m not afraid of writing less code, I’m afraid of caring less.