In the last piece the screen stopped being a thing someone shipped and became a thing the agent renders, fresh, each time I ask. I want to follow it one step further, into something I think follows from it but haven't really watched happen yet.
Let's start with something I have watched happen. When I ask an agent a question, most of the time I just take the answer. I could open the sources, see which sites it leaned on, notice that one was a careful publication and another was a forum post. I can, but most of the time, frankly, I don't. So most of the time, for most answers, I genuinely couldn't tell you where any of it came from. By the time they reach me they've all got the same voice, the agent's.
That has already happened to text. The thing I keep turning over is what happens when it happens to everything else. Because if the agent renders the screen, the way part one said it would, then the agent renders every screen. Same hand, same house style, every time. Run that forward and it leads somewhere I find quite bleak.
Everything starts to look the same.
Think about how you can tell one product from another without trying. The bank app that talks to you like a cautious solicitor, and the one that talks like a friend who's good with money. The checkout that feels expensive before you've spent anything, all weight and slowness, against the one that's all confetti and speed. The delivery tracker that's anxious and over-explaining versus the one that just tells you it's coming. A loading spinner you'd recognise in a lineup. None of it announces itself. You just know, the way you know your mum's handwriting, that this is them and not someone else.
All of that was a decision. Someone chose the blue. Someone argued for hours about whether the button should say Done or Got it. It looks like surface and it was never surface. It is the product telling you who it is in the only language a screen has.
Now picture the screen rendered by the agent instead. You ask for the thing and the agent draws you a panel. It's in the agent's type, the agent's spacing, the agent's colours, laid out the agent's way, because the agent drew it. Everything a product used to express itself with, the look, the wording, the small behaviours, the feel of the whole thing, never reaches you, because none of it gets rendered. Only the content does. Two products you'd have known apart at a glance arrive in the same panel, the same shape and palette and rhythm, separable only by the facts inside them.
Brand is never one thing you can point at. It's the accumulation. The colour and the type and the copy and the motion and the thousand small decisions about how the thing carries itself, all pulling the same way until they add up to a character you recognise. Take the rendering away from the product and give it to a neutral third party, and the accumulation has nothing to accumulate in. The character has nowhere to live but the surface, and the surface would belong to someone else.
At some point who has my loyalty is almost the wrong question. I wouldn't be able to tell who I was dealing with. They'd have dissolved into one interface, and the interface would belong to whoever built the agent.
It's the sources thing again, but for everything. The product is still there, still doing the work, but it reaches us only through the agent, and the agent is all we see. So the product becomes something like infrastructure. Present everywhere, noticed nowhere. Not disliked, just never met.
And even if a product wanted to hold on to its character, it's not clear what it would hand over. This is the part I find hardest to get around, because it isn't about taste, it's about who's deciding. When the agent renders the screen, it makes the design decisions, not the product. The product can send data, a price, a name, a colour as a number, but it can't send the choices, because the agent is the one choosing now. How it's laid out, what's emphasised, what the thing feels like to use, all of that is decided at render, every time.
Once the agent is the one choosing, what's left to tell two products apart is what they do and what they charge. Two things that do the same job for the same price are, through the agent, the same thing. So the work that used to go into being recognisable goes somewhere no one will ever see it, which is to say it stops being worth doing. At some point someone notices the design budget changes nothing about how the product reaches people, and spends it elsewhere. Why would they not.
I don't know how much of this is real and how much is me running a mood forward in a straight line. Brands kept their faces through the catalogue, the search results page, the comparison site, each of which was supposed to flatten everything into a row of near-identical entries and never quite managed it. Maybe character finds a way through this too, in a form I can't quite picture from here. I notice I want that to be true, which is its own kind of warning.
But I keep landing in the same place. The machines will draw us a clean, clear panel and find us exactly what we asked for, every time, and every one of them will be indistinguishable from the last. I just don't think they'll ever show us the thing we used to be able to see at a glance, which was that someone on the other side cared how this felt, and that they were not the same as everyone else.