should i just vibe code

So. A few weeks ago I was at the cafe (yes, the same one. variety in my life is famously not a thing) and I ended up randomly talking to a senior engineer for a bit. I don’t really remember what got said anymore. But something about that conversation set off a chain of thoughts in my head and the chain hasn’t really stopped since. So here we are.

The thing I keep getting stuck on is whether I should just fully start using AI for everything. Like for real. No more pretending I’m too principled for it.

Right now? Not really. I use AI for the small stuff. The blog. Little fixes. But for the actual coding work? I still mostly write it myself. I like going deep. I like understanding why things work. One of those dotfiles people. Or I’ve been one.

Lately though, something’s been shifting. And I don’t think I can keep this going.

A Month, A Day

So now I’m stuck with this math problem.

If I sit down and build a project without AI, putting in the hours, the days, the random side-quest thinking, it takes me a month. Someone else does the same thing with agentic tooling in a day.

A month. A day.

Why am I doing it myself.

And it’s not even just the big projects. Even my own dotfiles, I’ve started using AI to fix them up. Which is wild because dotfiles are supposed to be the one thing you do yourself, with full intention. But the thought process is just: what’s the point of spending hours on this when AI gets me there in five minutes?

The math problem isn’t theoretical for me. It’s already eating into the stuff I thought I cared about.

Like, the integrity. The dignity. The care. Yes, all real. I care about all of it. But at the end of the day, is that gonna get me that mf job I want?

I don’t think so. ;-;

So Here’s the Thing

I haven’t actually vibe coded a proper project and put it on my GitHub yet. I’ve used AI for shit like making this blog and other small stuff. Nothing I’d call a real agentic project shipped under my name.

(Also I think “vibe coded” is a bad way to call it anyway. Agentic engineering. Or whatever you wanna call it. Same idea, less of the meme baggage.)

But I’ve started one.

My first actual properly agentic engineered project, where I’m basically just a reviewer reading the code Codex produces. I review it. I steer it. I’m not the one writing it line by line.

And it has a really good tech stack. Terraform. EC2 Nitro Enclaves. Go. Stuff I have genuinely never touched before. And somehow? It all fucking works. I’m shipping faster than I ever have.

I’ve also been using the same approach for little open source contributions on the side. That part is fine I guess, just doesn’t hit the same. The satisfaction isn’t there. But the output is.

So what’s stopping me from putting that on my CV? Posting it on LinkedIn? A few more of these and I can portray myself as basically anything I want lmao. Is that where we’re headed?

The Part That Actually Gets Me

I’m leaning toward this. I could probably agentic engineer something solid in Rust or Zig within a matter of days. A really good project, even. Put it on the CV. Boom.

But here’s the part I haven’t said yet.

Even if I commit, the part of me that actually liked this just doesn’t come back. The going-deep part. The figuring-it-out part. That just goes.

There’s no getting that back.

ykwim?

That’s not it, gng.

So Now What

Pick a side or stay in the middle.

Use AI: actual results, no satisfaction. Don’t use AI: satisfaction, no actual results. The middle: doing both kinda badly. Which is basically where I’ve been this whole time.

But the way I see it now? There’s a game that needs to be played. And if you don’t play, I guess you get cooked. That’s kinda the end thing. At least that’s what I’ve been seeing.

idk.

Anyway. That’s where I’m at. Contemplating. Leaning. Confused. Hope my point landed. :))