Is vibe coding dying? · Automation-Tribe-Free

Is vibe coding dying?

You're probably still vibe coding everything and it feels great, right? We were doing the same thing six months ago. Shipped our entire MVP with v0 and Bolt. Three landing pages in a week. Felt unstoppable.

Then we tried adding payments. And user roles. And email notifications.

Every single feature broke something else. We'd build something in 20 minutes, then spend 3 hours fixing what it messed up. The codebase was a mess because we never planned the architecture, we just kept prompting and hoping.

Don't get me wrong - vibe coding is amazing for what you're probably doing right now:

Testing ideas fast
Building quick prototypes
Shipping demos to show investors

But here's what nobody tells you: the moment you need to scale or bring someone else into the project, it completely falls apart. Good luck explaining your vibe-coded structure to a new dev.

So we switched to spec coding.

Now we spend 15-20 minutes writing out what we are actually building. The requirements, how it should work, where things go. Then AI generates the code based on that plan.

Yeah, it's slower to start. But way faster to finish. Last week we created multiple landing pages with ongoing iterations every week - no bugs, no surprises, just worked as expected. Because we designed it before building it.

We still vibe code when we are just messing around with ideas. But for real features that need to last? Specs every time.

Most serious devs I know have made this shift. They realized the bottleneck isn't writing code anymore - AI does that. The bottleneck is knowing what to build and how it fits together.

Where are you at? Still in the vibe coding honeymoon phase or starting to feel the pain?

1 comment