Figma isn't enough anymore

Speed is one of the most desirable traits of any product or service.

How fast can it be done?

How quickly will I know?

How long do I have to wait?

Instant.

Especially if quality doesn’t suffer.

Fast and high quality is better than slow and high quality in most cases.

There are no doubt some caveats, but across the board this is true.

AI has dominated the market’s conversation because of its speed.

The quality depends on the task, but it’s getting better and better.

And it’s only getting faster and better.

Not worse.

That speed is already reshaping how companies are built.

Last week, Block, Jack Dorsey’s parent company of Square, Cash App and others just laid off 4,000 people out of ~10,000 employees.

He said…

Intelligence tools is a fancy way of saying AI.

This is what I want to talk about today.

AI-native design.

A term I believe we’ll be hearing a lot more of this year and one that will be a requirement soon for most companies that haven’t already shifted.

AI-native design means reaching for AI first and incorporating it as a natural part of your design and building process.

For a long time we had the “should designers code” debate. That point is now moot.

Prompting in natural language and having a rocket ship at the tips of your fingers changed everything.

But this doesn’t mean relying on fully AI-generated slop.

It’s not binary, it’s a continuum.

It doesn’t mean prompt once, see a failed attempt and declare “this sucks” and never use it again.

It means figuring out the most meaningful way to control the design and build process using the fastest tools we have available.

Speed.

With that in mind, here’s a starting point for you to experiment with and potentially incorporate into your own AI-native skill set.

I think of it as Code and Canvas, and at a high-level the process looks like this.

↑ That’s the process and what I strongly believe you need to embody to stay as relevant as possible this year and in this industry.

Step three is a wide continuum right now between Canvas and Code, but it’s no longer either/or.

Let me break down these 4 steps with a simple little exercise.

I’ll use an Activity Tracker, but this can obviously apply to whatever you’re working on, whether it’s a small UI card or a full blown project.

You can scope this accordingly based on what you’re working on.

Prompting first.

Here’s a format you can use as a starting point. If you don’t know what to write here, paste the format into an AI chat and say “interview me about what I want to design, and use this format”

In this example I’m using Figma Make, here’s my starting prompt.

From there I pulled into Figma for some Canvas exploration and ran some manual design experiments.

After coming up with some ideas, I copied the frame, and then pushed it back into Figma Make.

You can see that some details are captured, others are not.

At a certain point prompting the fine details in a tool like Figma Make feels too slow. Especially if your manual designs aren't getting read in correctly.

This feels like a natural place to download the code from Figma Make.

Then I put the code in a local folder on my machine, and open my terminal (iterm2) and run Claude Code.

This gives me MUCH more control and power vs only staying in Figma Make.

At any point during my Claude Code and browser experiments, I can ask Claude to send it BACK to Figma, using the new Claude Code → Figma capabilities with Figma’s MCP server.

We can also easily open this with Cursor as well, if you feel more comfortable not using the terminal.

I know this is new territory if you haven’t touched code before, but the core idea is simple, you’re sending a design back and forth between Code and Canvas. That's it.

Designing on Canvas (Figma, etc.) lets you move things around manually OR prompt.

Designing with Code lets you move things around with prompts only.

BUT...

you get the added benefit of it being a real living and breathing thing.

You can add rich animations, interactivity, integrate open source components, and more.

This is the new AI-native workflow.

Designers learning these techniques have a unique advantage to get really creative with these new super powers.

Hope this helps you in your designer journey!

PS: AI-native design is becoming (already?) the new baseline. Inside Shift Nudge, we’re currently building this into the curriculum. Code and Canvas, prompt-driven workflows, the whole thing. If you want to be ahead of this curve instead of catching up to it, apply now while this is being built.