Skip to content
Compound Design
5

State of Design 2026 · Part 1 of 5

The three tribes of vibe coding

The profession didn't migrate to AI as a block. It split into three groups that coexist on the same teams. This series starts with them.

The hook

The second most used tool among designers is an AI

82.6%
use Figma every week
50.8%
use Claude every week, the number 2 tool
48.2%
use ChatGPT every week
38.4%
use Claude Code, more than FigJam and Figma Make

I expected to see Figma at the top. I did not expect to see an agentic coding terminal ahead of FigJam in the daily work of people who do design. That's the data point that made me read the whole survey, and it's why this series exists.

The survey

Where the numbers come from

State of Prototyping Spring 2026 is a UX Tools survey of 1,478 designers and builders, collected between March 14 and April 6, 2026 across 18 regions. The data is open, under a CC BY 4.0 license.

The survey defines vibe coding honestly: “AI-generated code the designer may not fully understand, but that works.” It's not “using AI.” It's letting the AI write the code you wouldn't write yourself.

One caveat I'll repeat in every article: the sample is self-selected, drawn from the UX Tools newsletter and sponsor networks. People who answer tend to like tools. Read the adoption percentages as the profession's ceiling, not its average.

The core data

It's not an adoption curve. It's a split

How much of your output is vibe coding?
vibe_distribution · n=1,478 · “AI-generated code you may not fully understand, but that works”
None (0%) — 37.7% · n=55737.7%None (0%)Occasionally — 18.5% · n=27418.5%OccasionallyAbout half — 12.7% · n=18812.7%About halfMost of it — 17.5% · n=25817.5%Most of itNearly all — 13.6% · n=20113.6%Nearly allDon't use37.7% · n=557Complements31.2% · n=462Majority of output31.1% · n=459
View table
Tier%n
None (0%)37.7%557
Occasionally18.5%274
About half12.7%188
Most of it17.5%258
Nearly all13.6%201

Source: UX Tools, State of Prototyping Spring 2026 (CC BY 4.0)

If this were a classic diffusion curve, the middle would be fat and the ends thin. It's the reverse. 37.7% generate nothing with AI. 31.1% generate most or nearly all of it. Only 31.2% sit somewhere between “occasionally” and “about half.”

I gave the three groups names so I could talk about them: don't use, complements and majority of output.

Reading

Why the average hides what matters

When someone says “43.8% of designers already vibe code on half or more of their work,” the number is right and the reading is incomplete. It suggests a profession halfway through a transition. That's not what the distribution shows.

What the distribution shows is two ways of working coexisting, with a transition group between them. On the same team there are people who have never opened a terminal and people who ship a working prototype in an afternoon. The two look at each other strangely. Neither one is wrong.

For anyone leading a team, that changes the question. It's not “where are we on the curve.” It's “which tribes do we have here, and what does each one need.”

By role

Design Engineers at 80.9%. IC Designers at 35.0%

Who generates 50% or more with vibe coding, by role
vibe_by_role · % of each role · n in parentheses
Design Engineer — 80.9% · n=94Design Engineer (94)80.9%Lead / Principal — 56.8% · n=280Lead / Principal (280)56.8%Non-designer — 50.9% · n=57Non-designer (57)50.9%Manager / Director — 46.6% · n=161Manager / Director (161)46.6%IC Designer — 35.0% · n=862IC Designer (862)35.0%Researcher — 26.1% · n=23Researcher * (23)26.1%

* Researcher: n=23, directional only.

View table
Role%n
Design Engineer80.9%94
Lead / Principal56.8%280
Non-designer50.9%57
Manager / Director46.6%161
IC Designer35.0%862
Researcher26.1%23

Source: UX Tools, State of Prototyping Spring 2026 (CC BY 4.0)

The gap between Design Engineers and IC Designers is 46 points, inside the same organizations. That's not a difference in tooling. It's a difference in mandate.

Tension

The seniority paradox

If that reading is right, the barrier isn't technical competence. It's permission and time. And that's something a design leader controls.

Proportion

Where the IC fits in the story

IC Designers are 58.3% of the sample. Even with the lowest adoption rate among designers, they account for roughly 47% of all heavy vibe coders in absolute terms: about 302 of the 647 people who generate half or more with AI.

In other words: most of the people already working this way are ICs. But as a share of their own group, the IC is the slowest tribe among designers. Both sentences are true at once, and the second one is the worrying one. I come back to it in the third article.

Limits

What the survey doesn't say

Question

For the audience

Which tribe are you in? And your whole team, if you drew its bar?

What to do on Monday: draw your team's bar. Name the three tribes without judgment. Then pick one experiment per tribe: the people who don't use it get a small, safe task to try; the people who complement get a whole prototype to ship that way; the people generating the majority get responsibility for documenting how they review what the AI writes.

Continues

In the next article

If 43.8% already generate half or more of what they produce with AI, why do only 1.4% trust that output without reviewing it? The second article is about that distance, and about why the survey's name, State of Prototyping, is literal.