Your vibe-coded app is 80% done. The last 20% is design.

AI builds everything you asked for. Then it adds 20% more.

That’s the pattern I see in pretty much every vibe-coded product a founder shows me. The models keep getting better, and the old giveaways, like random gradients, are fading. But one thing hasn’t changed: the too-muchness.

Every idea the founder had is on the screen, plus a few the model added on its own. Nothing is wrong, exactly. It’s just all there, at once, at the same volume.

These screens are built to cover everything. They aren’t built for attention.

What does a vibe-coded screen look like?

We recently worked on a trading screen for a single instrument. Screens like that are full of data by nature: prices, volumes, positions, history, news, stats.

The vibe-coded version had all of it, stacked top to bottom. No structure, no layout, no hierarchy. Everything there was, put on the page, and hope for the best.

The data was right. The screen was unusable.

How do you fix it?

We didn’t start with pixels. We started with the content.

We went through every piece of data and asked the same few questions:

• Does it need a chart, or is a number enough?
• Does it need a title? A link?
• Can it sit behind a “read more”, or collapse?
• Does it belong in a tab?
• What’s the priority? What goes in front of the user, and what steps back?

Only once that was clear did the screen come together. Same data, much clearer picture. Something a trader could actually read at a glance.

That sorting is the design. The visuals come after.

Why not just prompt it yourself?

You can. Anyone can prompt.

The difference is what you prompt for. Founders prompt from the idea: this data needs to exist, so put it here. We prompt from a different place. How do we make this great? How does it read? How does it work with everything around it?

That angle comes from doing product design every day for 15 years. From knowing the foundations, and from looking at each piece of content critically before it earns a spot on the screen.

Founders prompt for what should exist. Designers prompt for what should matter.

Same tool, different questions.

Do you have to start over?

No. If you’ve vibe-coded your product and it works but doesn’t feel right, you’re 80% there. The last 20% is deciding what to leave out, what to bring forward and how it all fits together.

That’s the part we do at Bondin: a two-week design sprint, handed back as working code.