Taste Comes From the Craft

On 10 November, at Adobe MAX in Miami Beach, Adobe announced that its Creative Agent is expanding across Creative Cloud and Firefly. You direct it in your own words and hand off complex, multi-step jobs. Photoshop, Illustrator, Premiere and After Effects are getting local MCP integrations, so third-party agents can edit your native project files while layers and compositions stay live. There are agents in preview that keep working in the background after you step away from the desk.

The reasoning is stated plainly in the release. The agent takes the multi-step work so that creatives can "spend more time on the taste, style and vision" that makes their work stand out. Adobe's incoming chief executive says the same thing in his quote: the point of all of it is work that reflects your own taste, style and vision.

That is not a side benefit buried in a feature list. It is the pitch. Taste is the thing you are being freed up for.

And it matches what the industry has been telling itself for two years. The tools no longer matter; what matters is taste, and the decisions that come out of it. Taste is the part AI cannot replace, so the designer who has it will be fine.

I agree with the conclusion.

But where does taste come from?

Taste is not a gift

Taste arrives from doing the work. Not from looking at the work — from making it.

It comes from pushing the buttons. Moving the pixels. Drawing sketches that embarrass you. Getting it wrong a hundred times and starting again on the hundred-and-first. Every one of those attempts leaves a deposit, and the deposits are what you are drawing on later when you glance at something and know, without being able to explain it, that it isn't right yet.

That knowledge is not available any other way. You cannot read your way to it, and you cannot buy it with a subscription. It is accumulated residue from your own failures.

Which is why the position is unstable. If taste is the thing that survives automation, and taste is built by the labour that automation removes, then we are being sold a tool that frees us up to spend something we have stopped earning.

The proof is in your own portfolio

There is a simple test for this, and everyone who has worked long enough has already run it.

Look at what you made fifteen or twenty years ago. At the time, I thought mine was the cream of the crop. Logos, drawings, layouts — I was certain they were excellent. Now I look at the same files and see the mistakes immediately. Not subtle ones. Obvious ones, sitting there in plain sight, that I could not see at all when I made them.

Everyone knows that feeling. You believed you had made a masterpiece, and a few years later it is unmistakably not one.

That gap is the only hard evidence of taste that exists. Your eye moved. And the only reason it moved is that you kept grinding — kept making things, kept being wrong, kept noticing. Nobody handed you the correction. You produced it yourself by continuing to work.

Nobody else ever saw them

Here is the part that should worry us more.

Some of that work is still out there, still running. The client was happy with it then. The client is happy with it now. In all the years since, nobody has walked into that office and pointed at the mistakes, because nobody else can see them. I can. Only my own evolved eye can, and only today — I could not see them when I made them either.

So the mistakes were never caught by the market. They were never caught by the client, the stakeholders, the approval round, or the satisfaction survey. They were caught by one person, years late, privately, and the only consequence was that he stopped making them in his next piece of work.

That is the actual feedback loop the design industry runs on. Not client approval — client approval measures expectation met, not quality. Not metrics, which tell you something moved without telling you what good would have looked like. The loop that improves the work runs through a designer who keeps making things, and whose eye sharpens faster than his clients' does.

It is a single point of failure, and we have never acknowledged it, because for a hundred years it was never at risk. The designer was always making things, so the detector was always being built, whether anyone was paying attention to it or not.

Now fast forward. If the eye stops getting built, the detector goes quiet — and nothing else in the chain is positioned to replace it. The client was never going to see the mistake. The stakeholder was never going to see it. If we stop seeing it too, the work does not get worse in any visible way. It gets worse with nobody left to notice.

Slow work

There is a name worth giving to the part that is being handed off. Slow work.

Slow work is the donkey work. The hours of pushing pixels, redrawing the same shape, rebuilding the layout for the fourth time, grinding a concept into existence. It is unglamorous and it is slow by definition, and it is exactly what the multi-step handoff is designed to remove.

But something else happens during those hours, and it is not visible from the outside. While your hands are busy, you are thinking about the work. Not deliberately — it runs underneath. You notice the thing that doesn't sit right. You half-solve a problem three steps ahead without meaning to. You change your mind about the whole direction somewhere around hour four, for reasons you could not have reached by sitting and considering it.

This happens even when the concept turns out to be wrong. Especially then. The hours were not wasted because the output failed; the output was never the only thing being produced. The thinking was.

That is the hidden part. And it is what you lose when the concept arrives finished in your hand, instantly, with nothing behind it.

Fast is not the problem

This is not an argument against speed. Fast processes are good. Nobody should want the afternoons back for their own sake, and I use these tools.

The problem is that the slow work and the thinking were the same activity, and when you remove one you remove the other without noticing, because only one of them was ever on the invoice. You were paying for execution. You were receiving judgement as a by-product. Cut the execution and the by-product goes with it, silently.

Which makes it a hard loss to see. Nothing on your screen tells you it happened. The work still comes out. It comes out faster and it looks better. The only thing missing is the hours you would have spent inside the problem, and you cannot miss hours you never spent.

New things are made of old things

AI can make new things. That is not in dispute.

But if you want to make new things, you have to have made the old ones first — with your own hands. That is where the material comes from. Originality is recombination: you reach for something you did years ago, on a project nobody remembers, and you put it next to something in front of you now, and you see what happens. The idea arrives because your hands have been there before.

If you have not done the old work, there is nothing in the drawer to recombine. You can still direct. You can still describe what you want and judge what comes back. But the judgement is running on a fuel supply you never built.

The speed is the danger

What the agent gives you arrives finished-looking. And a finished-looking thing invites you to stop — to decide it is done, because it looks done and it took nine seconds.

That is the same error I made twenty years ago, looking at my logos. The difference is that I had to spend weeks building the thing I was wrong about, and those weeks were also teaching me. Now the error is available instantly, repeatedly, at no cost, and with no lesson attached.

Pointing AI at a concept you can already see in your head, and aligning what it produces against your own judgement, is real work and good work. The question is only what happens to the judgement over time.

The open question

So that is what I am left wondering, and I do not have the answer.

If the craft is where taste is made, how do we make taste now? Where does the next generation get its deposits from? What replaces the hundred failed attempts, and the slow unglamorous process of being wrong in public and noticing it years later?

And if the designer's eye was the only thing in the chain that ever caught the mistakes — what happens when we start missing them too? Who is left to tell us? The work will still ship. The clients will still be happy. That was always true.

Adobe is right that taste, style and vision are what will matter. I just want to know where we are going to get them, once the thing that built them has been handed off.

I used to look at my own work and see a masterpiece. Now the machine hands me one at first glance, before I have done anything.

What will we see when we look back at it in a few years?