Code Got Cheap. Taste Did Not

An engineer on my team told me he wants to kill designers. I laughed, because I understood the argument. If a model produces screens and flows in seconds, why keep someone whose job is to draw the thing and hand it over? Kill the role. Keep the builders.

He has it backwards, and the reason is worth more than the joke.

This is the best moment designers have had in a long time. Not because the tools got better. Because the wall between feeling something and shipping something finally got thin enough that taste can move at the speed of the product.

What the word used to mean

For years the job was clear. You owned the artifact. You made the screens look right, you wrote the specs, you sat in the reviews, and your status came from the file and from the deck that explained the file.

That model made sense, because production was expensive. Someone had to translate intent into something engineers could build without guessing, and the handoff was the product of the role.

Then production got cheap. Models write UI. A prototype that used to take a week takes an afternoon. What did not get cheap is the middle of the product. The empty state. The answer the system is not sure about. The paywall moment. The thing that breaks at two in the morning. That part still needs a human who cares how it feels.

The artifact is no longer scarce. The feeling is.

Titles stopped predicting the work

Inside the teams building AI products right now, the old functions are melting into each other. Engineering, product, design and data science stopped being separate ladders and became five shapes of work that cut across every badge in the room.

The prototyper invents constantly and ships almost none of it. That is the job, not a failure of it.

The builder takes one of those ideas and makes it real at production quality, fast.

The sweeper cleans up. Simplifies the system, removes what should never have shipped, makes the thing feel sharp again.

The grower takes something live and pushes it toward the version people actually pay for.

The maintainer owns a mature system and keeps it secure, fast and reliable while it scales.

Designers already show up in more than one of those. So do engineers and product managers. The title on your badge matters less than which of the five you can actually do this quarter.

The old word designer was a domain label. The new words are verbs.

Decoration was never the job

The engineer wants to kill designers because he still believes design is decoration bolted on at the end. If decoration is free, decoration is dead, and on that narrow point he is right.

Design was never the decoration. Design is the judgment about what should exist, what should never exist, and what has to feel trustworthy when the system is wrong.

Anyone can generate fifteen versions of a screen now. Almost nobody can tell you which three are wrong for a Japanese user who needs consensus before deciding, which one converts, and which one quietly destroys trust the first time the model is unsure.

That judgment is the feeling. Call it taste if you prefer. I call it the part of the job that cannot be moved into a prompt.

Engineers who can feel that are already designers in practice. Designers who can build are already engineers in practice. The people stuck in the middle, waiting for someone else to own the outcome, are who the joke is actually about.

Shipping got cheap, feeling got expensive

Here is what most of the threads about AI replacing design miss. Code used to be the bottleneck, so feeling was cheap relative to shipping. A company could afford a team that only felt, because making the thing real was slow and hard.

That ratio flipped. The person who can sit with a half built product, notice that the upgrade moment feels greedy, notice that the refusal state feels cold, notice that the onboarding is lying about what the model can do, and then change the product that same afternoon, is the person who wins.

One of my own apps had an upgrade screen I was quietly proud of. Clean, confident, well spaced. I opened it one evening on my phone, in the state a real user would be in, and it read like a hand out for money before the product had earned anything. That is not a research finding. It is a feeling, it took four seconds, and it was correct. The version that shipped that night converts better and asks later.

Under the old model that observation becomes a ticket, a debate and a sprint. It usually dies. Now it is an afternoon, and the person who felt it is the person who fixed it.

Designers who learn to build get both halves. They do not wait for a ticket. They do not lose the intent in a handoff. They carry the feeling all the way into production. That is the same argument I made when I wrote that the handoff is dead, and it is why AI did not make me faster. It changed what I am slow at, and what I am slow at now is judgment.

I live in Tokyo, where craft is not a mood board word. A mediocre product stands out here immediately and nobody tells you. They just stop coming back. That standard is a gift at a moment when volume is free, because volume without feeling is noise with better margins.

Collect verbs, not titles

So stop defending the title. Start collecting the verbs.

Early in a product, get dangerous as a prototyper and a builder. Invent, ship, kill what failed, invent again.

Once it is live, get dangerous as a sweeper and a grower. Remove the clutter, find the moment people actually pay for, fix the failure states nobody designed on purpose.

You do not need a famous logo on your resume to work this way. You need a bias for the outcome over the artifact, and enough build literacy that the feeling survives contact with reality.

I write code. I ship my own products. I still care how a screen feels in the first three seconds. Those are not opposite skills. They are the same job with the middle removed.

The people who only ship will lose to the people who can feel. The ones who do both will not need permission from either tribe.