Vibe Coding：机会还是干扰？

It's never been easier to make things. As AI models improve week over week, we're (treated?) to a never-ending supply of mind-blowing examples of what it can do.

But crazy projects aside, the off-the-shelf tools from OpenAI and Anthropic make software coding suddenly seem so easy. I've tried my hand at it—I've built a software dashboard to keep track of the mastermind groups we've put together in ​The Lab​. And it seems like every week, one of the creators I admire has committed to vibe-coding their own SaaS platform.

Is this the move?

As always, the answer is "maybe." But I'm mostly skeptical.

To augment my own thinking, I interviewed Rob Walling, the co-founder of Drip and a legend in the SaaS world.

Rob built and sold Drip, an early email marketing software, for an 8-figure exit. He did it in just 3.5 years and walked away with life-changing money. In addition to Drip, Rob created ​MicroConf​ (a community + conference for software founders) and ​TinySeed​, which has invested in hundreds of other SaaS founders.

So the guy knows software, OK?

He's not sold on vibe-coding as an opportunity for most creators, and he had a great analogy for why not:

"If I were to invite you over to my house—and let's just say you and I don't have any construction experience or any carpentry experience—you and I could build like a little tool shed. We'd just figure it out, and it would work. We could also then go build, probably build an outhouse, right? Just a small little structure.

But the moment it's like, 'All right, dude, help me build a single-family house and a two-story house and a, you know, commercial skyscraper'...it's the same thing with vibe coding and coding. You can build a little utility...but the bigger you get, the more complicated you get. And if you bring in a developer later, it's like having a five-story building with a foundation that's totally hosed and asking, 'Can't we just clean that up?'

There are several important points in that one analogy:

1. He's not saying there's no value or opportunity for vibe coding. Small utilities, especially internal tools with users on your team who aren't paying for it, can be really useful and totally buildable.
2. Most software tools that actually generate revenue from customers are closer to a residential or commercial building. Except they're also never done—to stay competitive, you always need to be improving them. And the weaker your foundation, the more technical debt you have.

To put a finer point on it, Rob told me:

"Being a part-time SaaS entrepreneur...it isn't really a thing. It just doesn't exist. If you're doing it part-time—you're half-assing it—someone will swoop in and eat your lunch."

This is my concern for all the creators I see spending a ton of time away from their core business to begin vibe coding a tool they think they can turn around and monetize. Yes, you can build a tool that functions—but software is intense. So intense that even after Rob SUCCEEDED at building and selling a SaaS, he told his wife he wouldn't be doing a software company again.

And let's say you do build something of value that people are willing to pay for. Software is a notorious breeding ground for copycats—and there are plenty of cats waiting in the wings looking for validated ideas to copy. Copycats with real software experience! They can replicate what you made and improve it fast.

I don't think most creators actually want to build software companies. I think they want recurring revenue, software has always been the crown jewel of that, and now it seems so attainable. And beyond that, I think there's something more insidious behind the push to vibe coding...

You've probably heard about the famous study where rats would press a button to trigger the release of dopamine—ignoring all of their other physical needs like food and water—to the point of exhaustion.

LLMs are dopamine buttons and we are the rats. Being able to type of a vague prompt and receive some clunky-but-functional app back is satisfying—in a way we haven't really experienced before. And while you're promptcrastinating by pushing the dopamine button, you're ignoring all the other needs of your business. That's a high opportunity cost.

So here's my takeaway: If you want to build a software company, do it. But go all in. You can vibe code a proof of concept, but you should probably still find a technical co-founder. It's far better to give engineers the super power of AI coding than rely on a vibe-coded app to give you a long-term software business.

Shiny object syndrome has never been easier to catch. But if you DON'T want to go all in on building a software company, direct that time and energy elsewhere.

It's OK to stay the course.

PS: Check out the full interview with Rob Walling here: