Designing for Pressure

For decades, design treated friction as a flaw. Fewer steps, fewer clicks, less effort: the frictionless flow as the summit of the craft. Every version of an interface was measured by how much work it managed to lift off the person. Artificial intelligence takes that logic to its conclusion: it no longer removes steps, it begins to remove thinking itself. It drafts, summarizes, decides, synthesizes. And the promise is hard to refuse, because no one misses tedious work. Removing effort feels, almost always, like progress.

But there is a point where that logic breaks, and neuroscience has begun to name it. The neurologist Richard Cytowic puts it plainly: we have a brain with real limits, metabolic and attentional, facing an environment that overwhelms them by design. The constant stimulation of screens exceeds an architecture that was never shaped for that volume. It isn't that the brain is frozen in the Stone Age, as the title of his book suggests; it's that culture and technology change far faster than biology can follow, and that gap in speeds leaves us chronically overstimulated. His prescription follows from his diagnosis: use technology less, protect attention, reclaim silence. And he is right about the half he sees.

Because there is another half. Not all effort is overload. Some friction is not the obstacle to value: it is the mechanism by which value appears. Reading a book and reading its summary convey similar information, but only one transforms you, and what transforms you is precisely the work the other saves you. If we remove all friction we don't get a healthier mind; we get an untrained one. And there the design question turns all the way around. It stops being how we remove effort and becomes which effort is worth keeping, and in what dose.

The dose makes the poison

There's an old word, borrowed from toxicology, that organizes this whole problem: hormesis. It describes something that sounds contradictory at first. A low dose of something that would be harmful at a high dose can become beneficial. Exercise is controlled stress on the muscle, and that is why it strengthens it. Fasting is a measured scarcity that triggers repair. A vaccine is a threat in miniature that trains the immune system. In every case, what adapts is not the absence of stress nor its excess, but the right dose, applied with control and followed by recovery.

Cognitive friction works the same way. And this resolves a contradiction that, if we don't see it, blows the argument apart from the inside. I am saying two things at once: that we need more friction, and that we already live overwhelmed by stimulation. Both are true, because not all stress is the same. The kind that adapts is acute, bounded, controlled, and followed by rest. The kind that harms is chronic, diffuse, without pause. The overstimulation Cytowic describes (the notifications, the infinite feed, the alerts) is exactly the second kind: load without recovery. The friction I'm defending is the first: the hard stretch of sitting down to understand something, which begins and ends, and which you come out of having learned.

Seen this way, Cytowic isn't wrong; he is describing only one half of a curve. His insistence on silence, which he calls an essential nutrient, fits perfectly on the other side: silence is the recovery phase without which no effort ever adapts. The muscle doesn't grow during the set; it grows during the rest. A good system shouldn't have to choose between demanding and letting breathe. It should do both, in rhythm.

For design, this changes the question. For years we measured success by reduction: less friction, less load, less time. Hormesis suggests the goal is not to minimize effort but to calibrate it. Not zero friction, which atrophies. Not infinite friction, which overwhelms. The dose in the middle, the one that makes you work without breaking you.

What extremophiles know

This principle is far older than any interface. It is, in a sense, how life works.

In the most hostile places on the planet, hydrothermal vents at temperatures that would cook any other creature, lakes so salty or so acidic they seem sterile, there are organisms called extremophiles. What's interesting is not that they survive those conditions despite everything. It's that they are made for them: their biochemistry was shaped, generation after generation, driven by that very pressure. The stress of the environment was not the enemy of their development. It was its engine.

At the scale of species, selective pressure carved out capacities that would never have appeared in a comfortable world. At the scale of an individual, plasticity does something similar in real time: we become capable of whatever demands it of us. They are two different clocks: evolution takes millennia, learning takes an afternoon. But they tell the same story, and that story speaks directly to anyone who designs. A tool that takes all the demand off us offers the cognitive equivalent of an environment with no pressure at all: comfortable, and slowly weakening. A capacity that goes unused is not preserved. It is lost.

It's worth not romanticizing stress: too much pressure forges nothing, it kills. But that, again, is the matter of dose. The point is not that suffering is good. It's that life builds its abilities by responding to a challenge, never in its absence.

We don't think alone

There is a second biological principle that changes the way we see this even more, and it begins from an idea we already know in design: we don't think alone. In an earlier essay I wrote that the notebook, not the AI, is often my real thinking interface. That has a name in philosophy of mind, the extended mind: the idea that thinking doesn't happen only inside the skull, but distributed across the person, their objects, and their environment. The notebook, the map, the search engine, and now AI, are part of the process; they are not external to it.

Biology has its own version of this, and it's more radical. It's called the holobiont: the idea that an organism is not an isolated individual but a consortium, a body plus all the microorganisms that live with it and without which it wouldn't function. It's not a gentle metaphor. The biologist Lynn Margulis showed that mitochondria, the structures that produce the energy of every one of our cells, were once free-living bacteria that ended up inside another cell and stayed to live there. That fusion, that ancient alliance, is what made complex life possible. We didn't get here by competing alone. We got here by integrating others.

If the extended mind is real, then each of us, while thinking, is a kind of cognitive holobiont: a person plus the tools they think with. And there the design question becomes the sharpest of all. When we bring AI into that consortium, what kind of symbiont are we adding? Because biology knows two. The mutualist, like the mitochondrion, which gives the host a richer life than it would have alone. And the parasite, which settles in, takes advantage, and leaves the host weaker than it found it. The same relationship of dependence can go either way. Much of the ethics of AI design fits inside that distinction.

Designing friction, and deciding the dose

What does it mean, concretely, to design a mutualist AI? It means no longer asking only what we can automate, and starting to ask what we should not automate. AI can generate a strategy, but it cannot generate the strategist: judgment is not delivered in an output, it is built by exercising it. Designing for that is designing friction on purpose, in the place where it forms rather than obstructs.

Some forms are surprisingly simple. A system that, instead of giving you the answer, sometimes hands you back a question: what are you assuming, what evidence would change your mind. A flow that won't let you move forward until you can show you understood what the machine produced; a small test, not as punishment, but as proof that you went through the work and didn't just delegate it. An assistant that shows its uncertainty instead of performing certainty, and in doing so pushes you to verify rather than trust blindly. Each of these gestures deliberately reintroduces the dose of effort that creates understanding.

But here we have to hold a discomfort rather than cover it up. Designing someone else's friction is also deciding for someone else. If I build a system that interrupts your consensus, or that stops you until you pass a quiz, I am exercising a form of authority over the way you think: I decide which effort is good for you, and when you get to move on. At the scale of one person it looks like care. At the scale of millions of people using the same system, it is the design of an entire cognitive culture, with everything that carries of power. Benevolent friction and imposition resemble each other more than we'd like. I don't have a clean way to separate them, and I distrust anyone who says they do.

Perhaps that's why the right question is not entirely one of design, but of relationship. For years we asked whether a tool made us faster. Then some of us began to ask whether it left us thinking better. But both of those questions are answered in an instant, and this plays out over the long run. A mutualist and a parasite can feel identical on the first day; they are told apart only after months, when one has left you more capable and the other more dependent without your noticing.

So maybe what we should ask of every system we design is not whether it solved the task, nor even whether it improved a decision, but something slower and more uncomfortable: after a year of using it, did it make us more capable or just more comfortable? Biology makes no promise that the alliance turns out well. A mitochondrion and a parasite begin the same, as guests who arrive. What decides which one they become is not the tool: it's the dose of effort we choose to keep, and who gets to decide it.

And who decides is, to me, the heart of all this. No one is going to come from outside to steer this toward its best version: most of the time it will be each of us with our own judgment, working it out as we go. It can feel like being left to fend for ourselves. But it's really an open space, and open spaces belong to those who design.

Like the philosophers who once took on the great questions of their time, today we're handed one that's quieter but no less deep: what happens to human thought when we begin to delegate it. Not everyone has to carry it; it's enough to become conscious of the question to stop designing on autopilot. That's our place, if we dare to take it. And I hope we take it like people who never forget that on the other side, always, there is a mind taking shape.