Scroll through any dev timeline right now and you'll see it:
"Built this in one shot." 🚀
A full app, a landing page, a SaaS dashboard, all from a single prompt.
And honestly? It's impressive. The tools have gotten that good. I use AI every day at work and on my own products.
But I've never shipped anything I'm proud of in one shot. And I've stopped wanting to.
The first draft is the easy part
AI is incredible at getting you from zero to something. A layout, a working component, a reasonable API structure. That used to take hours. Now it takes seconds.
But "something" isn't the goal. The goal is the right thing.
When I was working on the landing page for Clarly (a WhatsApp AI support tool we're building), the first AI-generated version was technically clean. Good structure, no bugs, decent spacing.
It also looked exactly like every other default SaaS page on the internet.
Nothing was wrong. It just wasn't ours. So I went back in. Changed the typography, broke the symmetric card grid, replaced the generic hero with an actual WhatsApp conversation, because that's what the product is. Every change had a reason.
That second pass is where the product actually showed up.
Two questions I ask before I ship
I don't consider something done until I can answer these:
1. What did I change, and what did I actually need?
Not what I prompted. What I needed. Those are often different. The prompt is your best guess at the problem. Reviewing the output is where you discover what the problem actually was.
If I can't explain what I changed from the first draft and why, I probably didn't think hard enough about it.
2. How will the end user feel when they use it?
The user doesn't care how many prompts it took. They care whether it's clear, fast, and feels like it was built for them. That only happens when someone sits in their seat and asks, "does this actually make sense?"
AI can't do that part for you. Not yet, anyway.
One-shot is a flex. Polished is a product.
I'm not saying one-shot builds are bad. They're great for prototypes, demos, and testing ideas fast. Use them.
But there's a difference between generating something and building something. Generating is output. Building is intent, knowing what you changed, why you changed it, and who it's for.
So next time you see "built this in one shot," the more interesting question isn't how fast?
It's what would you change now?
How do you use AI in your workflow: one shot and ship, or draft and refine? Curious to hear in the comments. 👇
Top comments (0)