Where we think design is going

Everyone building a design tool has a view of where design is going. Ours is short enough to write down. Everything Wonder builds comes out of a few beliefs about design and code, and it seemed right to share them with the people we build for instead of keeping them in a company handbook.

There are four.

Code is the source of truth

Everything designed in Figma is outdated the second it is created. A design file is a picture of the product, and the product keeps moving without it. An engineer changes a padding, a component gets a new state, a color shifts, and the file does not know. So every team ends up with two versions of its product: the real one, which is code, and the picture they design in. Someone gets the job of keeping the picture in sync, and even the best teams fall behind, because the problem is not effort. The problem is that there are two different systems, and they never talk to each other.

We think code is the only real source of truth, because it is the only version anyone actually uses. A design file is an opinion about the product. The code is the product. So a design tool should start from the code, let you change what is there, and put the change back where it belongs, instead of keeping a copy on the side. When design and code are the same thing, the gap has nowhere to open.

Design should be interactive, not static

For a long time designers could not build what they imagined. You could draw it, and then you had to explain it: here is what happens when you click, here is how it should feel. A prototype was a stack of frames with arrows between them.

AI agents changed that. Now anyone can build a working prototype in an afternoon, with real data and real motion. But how we share those ideas, how we collaborate on them, how we get feedback, none of that is solved. Prototypes live on someone's laptop. Teammates get a screen recording, or a link that dies next day, and feedback comes back as screenshots. You cannot judge a living thing from a still picture.

We think those prototypes belong on the canvas, side by side, where the whole team already looks. Not pictures of the idea. The idea itself, running. You open it, click through it and feel it before anyone has to explain it. Sharing one takes seconds, and feedback lands on the thing itself, not on a screenshot somewhere else. A prototype is still an opinion, even when it runs. A canvas should be a place where ideas run.

Designers should code, and engineers should design

For years, titles decided who was allowed to touch what. Designers designed, engineers coded, and everything in between was a handoff: a rebuild, a round of screenshots, a backlog of small fixes nobody had time for. That handoff is where products lose their quality.

AI took away the reason to keep the roles apart. A designer with an agent can code. An engineer with an agent can design. The best teams already work like this. Design on canvas is still where an idea starts. It is just not where the work stops anymore.

What does not change is who owns quality. Design owns how the product looks and feels. Engineering owns the code that ships. So an engineer's design is an idea, and a designer makes it production ready. A designer's pull request is a start, and an engineer polishes it before it merges. Who can do what no longer matters. Anyone can start anything now. What matters is who is responsible for quality. Designers are responsible for the design. Engineers are responsible for the code. That does not change.

Craft cannot be delegated to an agent

Design was never about drawing rectangles. It is about understanding a problem well enough to know what should exist, and then working it out through iteration: try something, look at it, throw it away, try again, until it feels right. A first screen out of nothing is the easy part today with AI. The iteration after it is where the craft lives, and it is what we care about most.

In a minute you get a screen that looks finished. But looking finished and being right are different things. A screen can be polished and still miss what the person using it needs, because nobody worked through the problem. The judgment, the taste, the hundred small decisions that make a product feel right, those still happen in a person's head and hands. Hand that part to an agent and you get something plausible that nobody loves.

We believe truly magical products need human designers putting their magic into them. So let AI take the boring parts, the first draft, the repetition, the setup, and let a designer take over at any moment and shape the rest directly. Grab the thing and move it, not type "move the padding four pixels to the left" and wait. Anything an agent makes, a person can take further by hand, and a tool that only works by prompting has delegated the wrong part. We measure ourselves on how good the product feels at the end, not on how fast the first draft appeared.

What these add up to

Put the four together and they describe a different job for designers, not a smaller one. Design starts from the real product. Ideas are judged running. Anyone on a team can start the work, and the people responsible for quality finish it. The drawing gets cheaper and the judgment gets more valuable.

These are the beliefs Wonder is built on. They are what we think is true, and they are what you will see in everything we make.