An agent can hand over a finished screen in the time it used to take to sketch one. It looks done. The spacing is even, the labels read cleanly, every state is there. And that is exactly the moment the thinking is most likely to stop, because the thing in front of us looks like it already had thinking in it.
The tool supplied the polish. What it cannot supply is the standard we were going to judge the polish against. Someone still has to decide whether the finished-looking thing is actually right, and that decision is not made by looking harder. It is made by having something to measure against.
That is the job of a design principle, and it is not the job most people give it. Principles get treated like inspirational posters on a wall: agreed with and ignored. The useful version is smaller, sharper, and practical.
A design principle is a decision tool. It does not tell us the answer. It tells us what a good answer has to satisfy.
It is what we reach for when the question is new, when nothing in the system covers the case, and something plausible is sitting on the screen waiting to be accepted or sent back.
A shared pattern answers a question that has already been settled. Use the pattern and move on. But every system has an edge, and the edge is where the work actually gets hard: the case the pattern did not anticipate, the tradeoff nobody wrote down, the new thing that has no precedent yet. A principle is what runs the edge.
The principles that hold up are not opinions someone had about design. They are compressed observations about how people actually perceive, decide, and act, and that is why they survive contact with tools nobody had when they were written.
Fitts's law is the clean example. The time it takes to hit a target depends on how far away it is and how big it is. True for a mouse pointer in 1980, true for a thumb on glass now, and true for whatever the next interface turns out to be, because it describes human motor behavior, not the technology in front of the hand. The principle outlives every pattern built on top of it. Hick's law and Jakob's law work the same way: attention and learned expectation do not care which decade built the UI.
That durability is why a principle is safe to hand an agent. The tool will change. The pattern will be rebuilt. The principle underneath it holds, so it is the one thing worth encoding for a system that has to keep making good calls after the people who set it up have moved on. We are not handing the agent our taste. We are handing it the observation the taste was standing on.
That does not hand the judgment to the agent. It hands over the floor. The agent can apply the principle and produce a sound default. Deciding which principle wins when two of them pull against each other, whether this case is the exception, whether the sound default is right for this audience and this moment, that is the part that still needs a person, and it is the more interesting part.
It is also why a principle is worth more than a rule. A rule tells a team what to do and stops working the moment the situation changes. A principle tells a team what to protect, which is exactly what holds up when nobody planned for the situation. A rule scales instructions. A principle scales judgment.
Judgment that lives only in someone's head does not scale, and it walks out the door when they do. A principle is judgment made portable. It lets a team make good calls without a senior person in every room, and it holds the bar steady whether a person or an agent did the work. Fewer expensive corrections, and a floor that does not depend on who happened to be reviewing that day.
I have done the opposite plenty of times. Argued a screen on instinct in a room full of people doing the same, and only remembered afterward that we had already written down the thing that would have settled it. The principle was not missing. I just did not reach for it, and reaching for it first is most of the discipline.
For a long time, judgment got spent on production, on a thousand small calls about spacing and labels and where a control should sit. The tools are taking that layer now, and it is tempting to read that as less room for people. It is the opposite. When the small calls are handled, judgment is freed for the heavier work: what is worth building, what good means here, what we choose not to ship.
A principle is what carries judgment to that work and keeps it there. It does not fence thinking in. It is the framework that lets thinking rise.