The Architecture of Attention

During my time in Peru I reflected on the systems that shape how we see the world. This essay traces the quiet shift from an internet that helped us compare reality to one that rewards reaction, and why building tools that restore context, transparency, and epistemic discipline has become necessary.

Type

Resource

Published

2026-03-08

Last updated

2026-08-21

Reading time

14 min

Some kinds of social beauty are made, not found. They live in the fragile work of people trying, together, to mean what they say. A scene, an institution, a shared language, a set of habits that keeps strangers honest enough to coexist. None of it is guaranteed. It can thin out quietly, like light leaving a street at dusk, until one day you realise you have been walking by memory. That shift became harder to ignore over the last few years, in the way the internet sounded when I opened it, in the way workplaces talked about “users”, in the way certainty became a commodity. It did not feel like a single collapse. It felt like a series of small endings I could not unsee.

By trade I am a product designer, and by temperament a builder. Systems appeal to me because they are honest. Given enough time, they reveal what they reward and what they quietly punish. The trouble is that people can live inside a system for years without naming its incentives. I did that too. It was easy to believe I was simply shaping interfaces, improving flows, chasing clarity. Easy, too, to believe that better design meant better outcomes, and that information, if organised cleanly enough, would behave.

For a while, I believed a version of progress that now feels naive. Not childish, exactly, but innocent in the way you can be innocent while still being competent. More access to information, I thought, would make us more informed. Online culture, however chaotic, still seemed to bend towards transparency. Institutions could be corrected by exposure. Hypocrisy could not survive documentation. Truth looked like a problem of retrieval and literacy, and if we built better tools for finding things, the rest would follow.

There was also a belief about myself, which is harder to admit. Moral cleanliness, I thought, could be preserved through adjacency. I could help design systems without inheriting the full weight of what they did to people. I could be the person in the room saying, “We should not do it that way”, without forcing a reckoning every time. Back then that felt like prudence. Now it feels more like a way of mistaking caution for innocence.

Before the drift

In my early relationship with the internet, it felt like a library that had learned to talk. Messy, yes, but alive. You could follow a thread, compare sources, find the original document, and feel the floor under your feet. Even when a forum was chaotic, there was an implicit respect for context. A link mattered. A quote needed a reference. If someone was lying, you could often point to the line where the lie began.

I never believed everyone was rational. I am not that romantic. But it did seem plausible that, given enough time and access, more people would build a better internal map of reality. The role of platforms, at minimum, was to lower the friction between a question and a decent answer.

Institutions made sense to me in roughly the same way. Not as saints, but as structures that, when pressured, tended to correct. Journalism had incentives. Politics had incentives. Corporations had incentives. Yet there were also norms. There were editors. There were standards that felt stubborn enough to outlast a few bad cycles.

To believe this is to live with a quiet trust. Not trust that everything is true, but trust that truth is, in principle, available. That there is a common world, and that it is possible to argue about it without burning it down.

Part of that trust rested on a simple assumption: most people are not trying to make things worse. They are trying to get by, protect something, belong somewhere, do what feels reasonable from where they stand. I still think that is mostly true. Which only makes bad systems more dangerous. Ordinary motives, routed through the wrong incentives, can still produce distorted outcomes.

Where it started to fray

The first crack I remember is not a headline. It is a sensation.

Late at night, phone glowing too brightly, I was doing that familiar thing where you tell yourself you are “catching up” on the world. A sequence of posts appeared, all technically about the same event, yet none of them seemed to describe the same reality. Images were cropped differently. Timelines were shuffled. Captions carried a moral verdict before they carried any information. In the comments, people were not asking what happened, they were rehearsing who they were.

What followed was something like vertigo. Not because I did not know what to believe, but because I could see how belief itself was being manufactured. The feed was not showing me the world. It was shaping my posture towards it.

None of this was wholly new. Media has always framed. Politics has always spun. But something had changed in texture. The velocity was higher, the confidence was louder, and the cost of hesitation had risen. Pause to verify, and you looked indifferent. Ask for nuance, and you sounded disloyal. “Let’s wait and see” had started to sound socially suspicious.

Months later, another scene. A meeting room, a screen at the front, a product discussion that should have been straightforward. We were reviewing a change that would increase engagement. It was not a lie, exactly. It was a pattern.

The pattern was simple: make the thing slightly harder to ignore, slightly easier to misread, slightly more emotionally sticky. Not enough that anyone could point to it and say “dark pattern” with confidence, but enough that it would add up. Someone asked whether the change would move the engagement metrics enough to justify the adjustment.

Everyone understood what that meant. The conversation was not sinister. It was ordinary. Commercially rational. The language was about lift, performance, efficiency, the familiar grammar of product and advertising. But the dashboard was still measuring proxies, attention, clicks, dwell time, short-term conversion, rather than whether the experience had become more genuinely useful to the person on the other side of the screen. In practice, the shortest route to those numbers often ran through salience and stimulation more than clarity.

Looking at the figures on the slide, I remember thinking: this is what the system rewards. Not necessarily deception, and not necessarily malice, but reaction. A design can be technically sound and still be pointed in the wrong moral direction.

There was no storming out. No speech. Just a nod, a concern raised, an alternative suggested. Then I went home, and the next day I kept working. That is one reason I do not allow myself to speak from a pedestal now. Complicity rarely arrives in theatrical form. More often it looks ordinary, socially acceptable, professionally legible. My own work lived inside the same attention economy I now criticise, and I benefited from it too.

The third moment was quieter, and it landed harder.

I was talking with someone I respect, someone smart, calm, not a stereotype. We were discussing a story that had been circulating, and I said something like, “I’m not sure that’s accurate.” They looked at me with a mild irritation, not anger, something colder. Then came the reply: “It’s true enough. You know what it means.”

True enough. The phrase was casual, but it carried a worldview. A narrative could become more important than the facts it claimed to rest on. The moral meaning of a story could be treated as more important than its accuracy. Epistemology, in other words, had become optional, so long as you were on the right side.

That night a sentence went into my notes app: I cannot outsource my epistemology.

Its full meaning was not clear to me yet. Only this much was obvious: something in me refused to keep renting reality from the loudest voice in the room.

The turn inward

None of the shift from observer to builder felt heroic. It felt uncomfortable, like waking up from a pleasant dream and realising you have missed a bill payment.

Design teaches you early that every interface is an argument. Every default is an opinion. Every metric is a moral choice disguised as a technical one. I had always known this intellectually. At some point, though, it became personal. “The media” stopped looking like a distant entity and started to look like a chain of incentives. Then came the less flattering recognition: I was part of that chain.

There was another admission to make as well, because without it the whole essay turns into theatre. My own mind is not immune. Outrage still pulls. Certainty still offers its clean little dopamine hit. The temptation to share something before verifying it still appears, because being first feels like being relevant. Lucid did not come from standing above those impulses. It came from knowing how strong they are, and not trusting the current defaults to make me better.

There is a particular grief in recognising that you have been shaped. Not brainwashed, not fooled like a cartoon character, but nudged, trained, rewarded. Looking back, the pattern becomes embarrassingly legible. The post shared because it felt socially rewarding. The article clicked because it confirmed what was already suspected. The quick judgement made because it was easier than sitting with ambiguity.

From there, the path splits. One option is cynicism, which is really just another form of certainty. The other is to build something that makes the harder behaviour easier.

That kind of building is not clean. A tool for navigating information inherits the fears, identities, and shortcuts of the people using it. It inherits the risk of being gamed, misread, or weaponised. It also inherits the question of who gets to define “credible”, and what happens when that judgement is wrong.

What I did not want was to become an arbiter. I did not want to build a new priesthood of “fact-checkers” with better branding. Something closer to a mirror felt truer, though mirrors are not neutral either. A mirror chooses its frame, its lighting, its angle. It can flatter, or it can humiliate.

So the real question became simpler and harder at once: what kind of mirror could be honest without being cruel, and useful without pretending to be God?

What Lucid is for

Lucid Report started, in practice, as mornings and nights. A laptop, a quiet room, the first coffee of the day, the cursor blinking like a dare. I would often work while my wife was asleep. Sometimes I would step outside afterwards and notice how the air smelled, salt or rain, and feel the odd contrast between the calm world and the frantic one on my screen.

In plain language, Lucid is an AI-powered news aggregation and analysis platform that tries to help people see a story as a cluster, not as a single performance. It gathers coverage of the same event from multiple publishers, then helps you compare what is being emphasised, what is being omitted, and how the framing shifts depending on the source. The point is to surface context, not just headlines, and to make it easier to ask, “What else is true?”

That is what it is. What matters just as much is what it refuses to be.

Lucid will not optimise for maximum time-on-site by making you angry. It will not hide sources behind a glossy summary that asks you to trust the machine. It will not pretend that bias is an anomaly, or that “objectivity” means the absence of values. It will not turn attention into a behavioural loop where the system learns which fears produce the best retention.

Those loops were deliberately left out, even knowing how effective they can be as growth mechanics. Train a system on what people click when they are activated, and you will get a system that produces activation. That is not an accident. It is the metric doing what it was hired to do.

People sometimes ask whether “balanced coverage” means neutrality. I do not think neutrality is the right word. Neutrality suggests the absence of stance, the clean hands of someone untouched by judgement. That is not real. Fairness is closer. Fairness means showing your work. It means being explicit about method. It means not pretending all claims are equal, while also not hiding disagreement.