Output Isn't Design

Every few months, a new tool promises to generate interfaces faster. Move words to product instantly. Collapse design directly into code. The pitch sounds appealing: faster output, less friction, design solved.

But what if the premise itself is backwards?

The Assumption Behind Every Design Tool

Karri Saarinen, CEO and cofounder of Linear, recently made this point sharply: new tools keep promising to generate interfaces faster, but they're built on a flawed assumption — that output equals design.

When a tool brags about turning Figma into code in seconds, it's treating design as a production pipeline. Input goes in, interface comes out. The faster that pipeline runs, the more "design" you get.

That's not design. That's reproduction.

Design Is Search, Not Production

Saarinen's argument cuts deeper than tool criticism. He frames design as search, not production — you start with a messy problem, and early on, you don't know the answer. The process of exploring, questioning, and refining intent is the design work.

This explains why architects still sketch by hand. Why designers prototype in code only after exploring in loose sketches. Why the best product decisions come from heated debates about what to build, not faster handoffs between tools.

The creative process doesn't compress. It needs the messiness, the back-and-forth, the uncertainty. Tools that skip that phase don't make design faster — they make design shallower.

What This Means for AI Tools

Here's where it gets uncomfortable for AI agent builders (including us): the same assumption haunts our space. When we promise that AI can design your app from a prompt, we're selling the output, not the thinking.

The real opportunity isn't making design faster. It's helping teams spend more time on the questions that matter:

• What problem are we actually solving?
• Who is this for, and how do they think?
• What does success look like, and how will we know?

These questions don't have prompts. They have answers that emerge from exploration, not generation.

The Takeaway

The next time a tool promises to "generate" design, ask: generate what, exactly? A faster handoff between stages, or clarity about what you're building and why?

Output isn't design. Intent is.