A few weeks ago, I watched a student do something that would once have taken a small team a full sprint. It took her an afternoon. She described the tool she wanted, talked to a coding agent for a while, corrected it twice, and by the end of the session, she had a working interface that fetched real data and behaved like a product. Not a clickable mockup. A thing that actually ran. She seemed almost embarrassed by how fast it had happened, as if the speed itself were evidence of cheating.
The part that stayed with me was not the speed. It was that the screen she had drawn, the interface, felt like the least important thing in the room. The hard, valuable work had moved somewhere else, and none of us in the room could quite name where.
I have been chewing on that question while reading two writers who, on the surface, have almost nothing to do with each other. One writes about markets. The other writes about applied AI research. This fortnight, they happened to describe the same shift, from opposite ends, and the shift is one that anyone who designs and builds for a living should be paying close attention to.
The market read: the interface stops being the moat
Linas Beliūnas writes a newsletter about where finance meets AI, and he spent a recent issue picking apart a strange event: a batch of free text files wiping roughly $285 billion off software stocks in four trading days. The files were a set of open-source plugins, mostly structured prompts and configuration, that let a general-purpose model do work that specialized software used to charge a fortune for. The legal plugin that triggered Thomson Reuters’ worst day on record was, in his telling, around 2,500 lines of prompt instructions.
His point is not that software is dead. It is sharper than that:
“For two decades, SaaS bundled three things: data, logic, and interface. The interface was the moat. MCP Apps unbundles the interface. The app becomes a backend.”
That sentence stopped me. The interface as a moat is the world I have spent my career inside. It is the thing designers were hired to defend. If you can drag an Asana task or edit a Figma frame without ever leaving a chat window, then the carefully designed surface that used to hold the customer relationship becomes plumbing. Linas thinks Wall Street is pricing the death of SaaS when it should be pricing the death of the middleman. I think he is half right in a way that matters for us specifically: the layer most at risk is the layer many designers have treated as the whole job.
The research read: the AI becomes your reader, not just your tool
Ethan Mollick comes at it from inside the work. In his latest piece, he announces a new book, and almost in passing, he describes building its website by prompting a coding agent for a few minutes instead of fighting templates for hours. That alone is the student in my studio, again. But his real observation is about what the relationship with AI is turning into:
“When the AI is no longer just your assistant, but your reader, your critic, and the gatekeeper standing between your work and its audience.”
He even confesses to cutting his em-dashes (a detail I enjoyed, given my own allergy to them) because they now read as a human tell. The deeper move is that he stopped trying to please readers alone and started designing a version of his book’s site for the AIs that will decide whether to recommend it. The interface is no longer the place where a human meets your work. It is increasingly the place where a machine evaluates it on a human’s behalf.
Mollick’s framing for all of this is co-existence rather than co-intelligence. The tool is sometimes better than you and sometimes hilariously worse, and the work is no longer a single problem you solve once:
“It is a relationship you negotiate, and re-negotiate, as the models get ever better.”
Where the two readings meet
Put the two side by side, and they are describing one thing from two angles. Linas sees the interface being unbundled from the outside as a market repricing. Mollick sees it from the inside, as a working writer discovering that the surface of his work is now negotiated with a machine. The screen, the thing we drew, is losing its monopoly on the moment of contact.
This is the part where the industry reaches for its tired binary: AI either makes designers irrelevant or makes them gods. As usual, that is the wrong question. The honest reading is more uncomfortable and more interesting. A specific kind of design work, the execution of surfaces, is being automated and unbundled fast. And a different kind of design work is becoming the only thing left worth paying for.
Here is the perspective I have landed on, and I hold it with some conviction. If the interface was your moat, you are in trouble. If the interface was your craft but never your point, you are about to matter more than you have in years.
Because once you unbundle the surface, what remains is exactly the work I keep arguing the design engineer is built for. Not drawing the screen, but designing the moment a person decides whether to trust the thing on the other side. When to hand the keys to an agent and when to refuse its help, even when it is offering. What a system has to reveal about itself in the first thirty seconds before anyone risks relying on it. How trust accumulates over six months, and what quietly destroys it. None of that is surface. All of it is design, and almost none of it can be expressed in 2,500 lines of prompt.
The student in my studio built a running interface in an afternoon. What she could not generate in an afternoon was the judgment about whether the thing should behave that way at all, what it owed the person using it, and how it should fail when it inevitably did. That judgment is the moat now. It always quietly was. The unbundling just made it impossible to keep pretending the screen was the valuable part.
So the question I am sitting with, and the one I would put to you: if the interface is becoming a backend and the AI is becoming the reader, what is the smallest, most human thing in your work that a model cannot negotiate away? Find that. Build outward from it. I suspect it is closer to the center of the craft than most of us admitted when the surface was still the moat.
Sources
Linas Beliūnas, Software Ate the World. Is AI Eating Software Now? — https://linas.substack.com/p/aieatingsoftware (and the companion piece on MCP Apps, AI Killed the UI — https://linas.substack.com/p/aikilledui)
Ethan Mollick, Co-Existence and the End of Co-Intelligence — https://www.oneusefulthing.org/p/co-existence-and-the-end-of-co-intelligence (with context from The Shape of the Thing — https://www.oneusefulthing.org/p/the-shape-of-the-thing)