I have never designed anything in my life.

I'm very unartistic, and I just stink at all of that.

So when I started using Claude Design, I experimented a ton just to figure out what I even liked…

• I asked questions
• I tried things 10 different times
• And I hit my limits fast without creating anything good

Ever had a similar situation?

Chances are, you made the exact same mistake I did…

You opened the best AI tool before you knew what you wanted.

But, as you probably already know, every AI tool has a usage limit.

Some have limits you'll never touch. Others run out in an afternoon.

So the second you start experimenting inside the expensive one, you're burning through your limits on the best AI tool just to work out what you want.

Which is why I experiment where it's cheap, and only create where it's expensive.

Here's what that actually looks like in 6 moves, from the first idea to the finished job.

1. Do all your exploring in the AI tool with the highest limits

For the first phase, pick whichever tool from your stack that you can run all day.

And make it the only place you experiment.

That way you don't have to worry about running out of usage, so you can explore and try interesting things.

In case you’re curious, for me that tool is ChatGPT (but use whatever you've got).

2. Collect other people's work before you make your own

Whatever you're making, go find 4-5 examples of it on the web and save them.

The good thing is that you don’t need any taste for this part.

You just need to know where to look:

• Designing something? Scroll a social platform like X or Pinterest and screenshot the looks you like
  ​
• Building a tool? Screenshot the app you're replacing
  ​
• Setting up an agent? Save other people's skills and rebuild them to your preference

3. Ask the cheap AI tool for multiple options

Now attach your screenshots, describe what you're building, and ask for a whole sheet of variations at once.

The point of this step is not perfection. You're just looking for a direction.

It's like reading a menu. You don't really know what you want until you see what's on it.

And when the AI builds out 4-5 versions of the thing you want, your chances of getting what you like go way up.

4. Run 5-6 rounds of refinement

Pick the 2 variations you like the most and ask for more like those.

Then do it again. And again.

I highly recommend getting to 5-6 rounds of this picking and expanding, because every refinement is one micro step in the right direction.

Btw, in this step also ask for a few options in each round (so the model doesn't lock onto one direction too early).

5. Have the cheap AI tool write the prompt for the expensive one

Before you switch to the expensive AI tool, ask the cheap one to write you a prompt describing every decision you just made.

AI is the best tool in the world for writing prompts (so don’t build them yourself).

Use something along the lines of:

6. Open the expensive AI tool and one-shot the whole thing

Now paste that prompt in, attach what you picked, and let it run.

If you did all 5 steps above, it'll take 1-2 turns instead of 10 (because every big decision is already made).

And you don't hit your limits.

Btw, if you're building something right now, come do it with us instead of on your own.

I run a live session inside the Vibe Coding Academy every Friday, where 800+ members are shipping their own apps.

Bring whatever you're working on, and we'll build it together.

​Join the Vibe Coding Academy here.​

Keep shipping,

Alex