arrow

first try at vibe coding


Vibe coding feels like casting spells more than writing software.

I've been using Cursor the past couple days to spin up a proof-of-concept app. The speed at which it scaffolds things is honestly bewildering. I can hardly keep up.

It’s changed how I write code. I’m leaning way more on conjurer than scribe — throw code at the wall, see what sticks, let the AI clean it up. Just write and see what happens.

Not gonna lie: it feels like magic watching the IDE fill in the blanks. I got a basic CRUD app working with a live database in just a few hours. Wild.

But there’s a catch: I don’t actually understand most of the code.

Since I haven’t been deep in the logic, I’m missing the context. I can ask the AI agents to fix what they break — and so far, that works — but I know I’m skating on thin ice. I can feel a reckoning coming. At some point, I’ll hit a critical bug the agents can’t patch, and I’ll be on the hook to dive into code I barely know.

Is this better than the old way? I think so... probably. But it’s also not the same game anymore.

Instead of building a polished MVP and evolving from there, maybe the new loop is: build fast, break it, rebuild better. Then do it again.

That feels like the shift. Vibe coding changes the rhythm. It’s not about shipping solid code once — it’s about cycling through drafts until something finally sticks.

It’s faster. It’s more experimental. I don’t have to think as hard — which is a relief. But that ease comes at a cost: I lose the deep understanding that used to come from wrestling with every line.

Maybe vibe coding is just the beginning — not of faster software, but of faster forgetting.

image


May 17, 2025

6:46PM

Alameda, CA