Iñigo Medina argues that software interfaces will not vanish, but accrete, and that companies treating AI as a fashion cycle are repeating old mistakes.
The screen is not going away, Iñigo Medina says, because nothing in internet history ever fully disappears. New paradigms arrive, old ones linger, and the result is not a purge but a thicker, messier stack of tools, habits, and protocols. That is why he is skeptical of the current talk about the end of interfaces, and why he thinks companies are mistaking a structural shift for a passing craze. The real test, he argues, is not whether AI looks inevitable, but whether organizations can stop behaving like adolescents chasing the next fashion.
The new software buzz, Cristina Alcalzamora argues, is not about the death of interfaces but about where the interface moves. Salesforce, she said, is building for interfaces that are not for users, while Notion is pushing deeper into an invisible layer that sits between the person and the machine.
I read the CEO of Salesforce saying that the world of software is changing, and they are developing the product for non-user interfaces. They are going much more toward the agentic layer, and what they want is to invest there because they believe that is where the future of software is.
2:50
All the latest updates they are making are focused on these more agentic layers, not the user layer, but an invisible layer, let’s say. You can go deeper with just a prompt, and then they give you a graphical interface, a table, or other components that arise from that message.
4:23
She frames the move as a claim about the future, but her own examples are more cautious than the slogans. Big platforms with stable user bases can afford to experiment above the familiar interface, while smaller products still have to keep ordinary users, data entry, and basic trust intact or the promised agent layer has nowhere to stand.
Maybe there is a future where we do not have to interact with those software systems from the 90s to consume the information they hold. Probably we will evolve toward voice systems, even emotional systems, so that the agentic layer knows the user personally.
6:01
Medina pushes the same idea further, but without pretending the transition is painless. New systems may change how people approach software, yet he keeps returning to the old problem of adoption: tools that are hard to use, slow to roll out, or badly governed do not become obsolete just because a pitch deck says they should.
Medina’s core claim is simple: digital history does not reset. Each new layer arrives with the old one still attached, which is why he doubts the fantasy that interfaces will simply vanish and leave a clean slate behind.
The screen as a way of relating to things is very hard to make disappear. There may be other kinds of interfaces, voice or an API interface, that do things, but what happens is that everything accumulates.
14:09
If you look at the history of these last 60 years, nothing dies, the proportions change. Mobile had a lot of weight, but mobile did not kill the desktop.
16:28
He rooted that skepticism in a longer story of interfaces. Browsers did not erase mail, FTP, or SSH; mobile did not erase the desktop; smart TVs did not kill the habits they were supposed to replace. The pattern, he argued, is accumulation, not replacement, and the result is a heavier cognitive load rather than a purer medium.
What happens is that we live with the burden that nothing ever disappears. When a new paradigm arrives, it does not kill the previous one.
14:50
That argument runs against a familiar Silicon Valley habit of announcing the death of one interface every time another gets fashionable. Medina treated the current AI moment the same way he treated the early web and mobile: as a genuine change, but not a total break. Even the supposedly obsolete terminal, he said, is back in use because products built around LLMs* are giving it fresh purpose.*
Internet, in Medina’s telling, does not resolve tension. It exports it, multiplies it, and leaves organizations to live inside it. The web pulls toward decentralization, he argued, while companies pull toward centralization, and the result is not harmony but a permanent strain that only has to be managed, never erased.
It is a strange medium in many ways, and it contains a germ that keeps forcing things toward decentralization and toward constantly rethinking them.
20:55
If reconciliation means eliminating the tension, getting rid of it, that is impossible. It is a tension like electricity, a tension you need in order to work.
22:50
Medina framed the internet as a rare medium because it created both an information economy and a set of cultures that do not fit neatly into older corporate logic. Open source, hacker culture, and the habit of sharing work without immediate payment all grew inside a system built on copying, reproduction, and low marginal cost. That history, he argued, is why the internet keeps pushing organizations to rethink authority, access, and who gets to participate.
It generated the information economy, it generated its own cultures, the open source culture, the hacker culture, the culture of people who work without expecting anything in return.
19:44
In Berners-Lee’s dream, the web was to be a decentralized space in which each person from anywhere on the planet would theoretically have the same access, the same rights, the same ability to reach information as others who traditionally had more access.
20:31
Medina’s answer to panic is not calm, exactly. It is education, understood as a habit of reading history, papers, and the adjacent fields that are usually ignored when a new wave starts looking like a total reset. In his telling, fear comes from mistaking the latest surface chatter for the whole story.
My recommendation, or what I always do, perhaps it sounds a bit naive, is this: educate yourself, educate yourself in the broad sense. The first thing you have to do is read, know history, and if you read a paper every week, you start to build a broader view.
25:32
The best way for that wave not to catch you cold is to keep making these broad readings of the subject. Education is the antidote to all that, because it keeps you from being carried away by panic.
30:25
He argues that LinkedIn posts, for all their value, flatten the world into a single surface. Academia and slower reading, by contrast, provide the missing depth: historical comparisons, labor-market context, and the reminder that every “revolution” has predecessors.
That argument lets him puncture the idea that the current AI wave is unprecedented. He points to earlier cycles, from motor cars to the industrialization of the United States, to show that job displacement has often been followed by retraining and rearrangement rather than simple disappearance.
When the first nodes connected in the 60s and they saw that they could use a protocol to serve information, that already seemed revolutionary. And the computers of the 50s that processed certain things were revolutionary too.
32:22
Whenever you have the word revolution in your mouth, it should be clear that it looks more like a mathematical series, N1, N2, N3, than something that has just happened. That is education for me.
33:22
Medina’s warning is not really about AI. It is about the social reflex that turns a technical shift into a badge of belonging, then calls the result strategy. Companies, he argued, often adopt new tools the way teenagers follow fashion, fast and a little defensively, until the trend starts to feel like common sense.
Organizations fail habitually because they behave like teenagers following a trend. That is my professional experience, and what I see in technology, design and data colleagues who write every day about fear of losing their jobs.
32:30
The antidote is education. You have to look at it as one element in a series, not as something that has just appeared.
33:11
His case rests on a larger claim: most organizations do not fail because they lack enthusiasm, but because they lack historical perspective. Medina says people inside companies hear every wave as a revolution from nowhere, when it is really one more step in a longer sequence of tools, protocols and expectations. Read that way, panic is a form of amnesia.
When you connect the first nodes in the 1960s and see that you can use a protocol to serve information, that already felt revolutionary. The 1950s computers that processed certain things were revolutionary too.
33:11
He extends that argument into the way firms try to introduce AI from the top down. Inside many companies, he said, innovation becomes a separate ritual, pushed by managers who do not fully understand what they are pushing, then slowed by the fear that has spread from outside. The result is a lot of pilots and very little change.
The people trying to drive it do not know very well what they have to drive, and sometimes they are looking for innovation for innovation’s sake. In those cases, it is quite difficult.
35:33
By the end, Medina had narrowed the argument to a blunt constraint: the bottleneck is compute, and the price of using tokens keeps the whole promise of scale from becoming frictionless. He did not deny that the stack is shifting, but he refused the easy conclusion that the shift makes everyone a builder or makes software itself disappear.
No. I don't know about that. If we read these plan changes as a sign that these companies are adjusting because we have started to break the bank, maybe. But if we look at the historical series of human inventions, if it is not the case, it will be.
46:22
The capacity of computation is something you can end up reaching. If it is a matter of limits, we have almost always found a way past them, which does not mean we do not have that problem now, but maybe five or ten years later we will have invented something that goes beyond those limits.
46:29
Medina’s reading of the market is less apocalyptic than procedural. Some software will survive because it saves tokens, some will be replaced because a team can now build its own tools, and some categories will simply get squeezed as agents consume far more compute than anyone had planned for. He suggested that the biggest surprise was not the rise of agents themselves, but how fast they became hungry enough to turn economics into the real boundary.
The thesis that everyone is a builder has no weight if we look at the historical series. It is true that the digital medium has created more builders, but that does not mean everyone is going to be one.
47:40
The thesis that everyone builds their own software makes no sense either. The more plausible one is that we look at the layers that are easiest, because WordPress did not make everyone a builder, but it did create a natural market around it.
48:36
That is where Medina brought the argument back to older software history. WordPress did not erase previous systems, he said, and neither will AI turn every company into a workshop of internal builders. What it may do, in his telling, is create a thicker market around the easier layers, where smaller firms build on top of larger ones instead of recreating every system from scratch.
Will AI make interfaces disappear?
No, Medina argues they will multiply instead. He says the screen, voice, APIs, terminals, and search will coexist, because digital paradigms usually accumulate instead of killing each other.
Why does he compare organizations to adolescents?
He says many firms chase AI as a fashion, then reverse course once the mood changes. In his view, that creates incoherence, fear, and half-built projects.
What does he think managers should do now?
He says they should touch the material, not just talk about it. That means using the tools, understanding the stack, and anchoring experiments in real business needs.
Does he think everyone will become a builder?
No. He says the historical pattern is that more people can build than before, but not everyone should or will, because organizations still need specialized products and services.
AI-assisted summary of Dcycle's podcast, verified against the original transcript.