Addicted, and My Audience Is Half a Million Robots

Ibrahim Fawakherji — ArchUp

I have been asking myself a question recently that has no comfortable answer.

Do I love architecture, or have I become addicted to it?

The distinction matters. Love is something you can set aside when the day ends. Addiction follows you into your vacation, into your reading, into the way you look at a city from a taxi window at six in the morning when you should be thinking about something else entirely. It restructures your perception without asking permission. You stop seeing buildings and start reading them: the economy behind the facade choice, the political moment embedded in the street width, the human psychology encoded in the distance between a door and the nearest window.

I am an architecture addict. I have been for years. I do not consider this a problem worth solving.

What I have been unable to find, despite considerable looking, is another person who lives this particular addiction in the same way. Not someone who works in architecture. Not someone who respects it or appreciates it professionally. Someone for whom it functions the way it functions for me: as an obsession that connects to everything else, that cannot be fully separated from economics, from history, from politics, from the behavior of cities under pressure, from the question of what built environments actually do to the people who inhabit them over time.

A person whose judgment I trust told me recently to find a community that resembles me. The advice was well-intentioned and I took it seriously. I went looking in the most obvious place: architectural podcasts and broadcast programs aimed at practitioners. I found several that had developed real audiences, particularly those focused on the business of running an architecture practice, on entrepreneurship within the profession, on the financial and operational realities of running a small firm.

I started listening. The beginning was genuinely satisfying, the specific pleasure of hearing people use the same vocabulary you use internally, of recognizing a framework someone else has developed independently for a problem you thought was yours alone. One of the hosts made an observation about AI that I found precise and honest: that AI systems are structurally incapable of saying they do not know something. They will produce an answer to any question posed to them, regardless of whether the underlying knowledge is present or invented. The confidence is uniform regardless of the accuracy. This is a real and underappreciated limitation, and hearing it articulated clearly felt like genuine intellectual exchange.

Then the conversation turned. Not sharply, but gradually and with the particular smoothness that characterizes transitions that have been practiced many times. The discussion moved from ideas about the profession to discussion of the firm’s service offerings, then to an upcoming course, then to a consulting package, then to a membership tier, then to a promotional offer for listeners. The content that had drawn me in was functioning as an audience for a sales conversation I had not agreed to enter.

I want to be careful about what I am saying here. There is nothing wrong with professionals offering services. There is nothing wrong with building a business around expertise. The people doing this are operating rationally within the incentive structures available to them, and some of them are producing genuinely useful work alongside the commercial content.

What I recognized, listening, was that my own practice cannot operate this way. Not because commercial activity is beneath criticism, but because the critical position I have maintained for more than six years at ArchUp depends on the absence of that particular conflict of interest. The moment a review becomes a potential sales lead, the review changes in ways that are not always visible but are always present. I have accepted this as a constraint on the work rather than fighting it. It means the work does not generate the revenues that service-oriented content generates. It also means that when I criticize something, the reader has reason to believe the criticism is not serving an ulterior commercial purpose.

This is not a moral stance. It is an editorial one.

When I look at the numbers that describe ArchUp’s actual readership, the picture is genuinely strange. Over five hundred thousand non-human visitors arrive at the platform’s servers within a twenty-four hour period. These are not readers in any conventional sense. They are indexing bots for search engines, training crawlers for AI systems, archival agents for various platforms, automated processes that read and classify and store the content without any human attention involved in the encounter.

Every article I write is being consumed primarily by machines. The AI systems that practitioners now use daily were trained in part on content produced by publications like this one. The knowledge that gets returned when someone asks an AI about a topic in architectural criticism or urban planning or construction economics is drawn from the aggregate of what people like me wrote before the AI existed to use it.

I find this simultaneously absurd and clarifying. Absurd because the image of a writer and his audience of half a million robots defies every conventional framework for thinking about what publishing means. Clarifying because it removes any pretense that the work is primarily about reach or influence in the ordinary sense. The robots do not respond to the writing. They do not find it interesting or boring. They do not tell other people about it or choose to return because a particular piece moved them.

The human readers are a smaller number. But they are the only ones who matter in the sense that the writing is designed to address.

The reader I write for is not a demographic. It is a condition. It is the person who looks at a building and cannot stop at the surface, who reads the site visit the way I read mine, who finds the connection between a curtain wall specification and a municipal bond rate genuinely interesting rather than tiresome. It is the architecture student who writes to ask about something that worried them in a design review. It is the practitioner who disagrees with a position I have taken and explains why with enough specificity that the disagreement produces something. It is occasionally the person who has no professional connection to the field but who has found in architectural criticism a way of thinking about the built world that reframes what they see when they move through it.

These people are my audience. The robots are my data.

A note on what this means for anyone who wants to engage.

I will disagree with you. Not as a performance of independence, but because the things worth discussing in this field are genuinely contested, and positions that cannot be challenged are not positions at all. If I write something critical about a project or a practice or an approach, the criticism is not personal and it is not permanent. It is an honest reading of what I saw at the time I looked, offered without the obligation to soften it for anyone’s commercial benefit.

What I will not do is use the criticism to lead you toward a service, a course, a membership tier, or a consulting package. The work is the work. If it is useful to you, I am glad. If it is not, there are other writers.

But if you are also addicted, if architecture functions for you the way it functions for me, as a compulsion that connects to everything else and refuses to resolve itself into a stable professional transaction, then I would genuinely like to hear from you. Not to agree. To compare observations about the same phenomenon from positions that may have nothing in common except the obsession itself.

The email is in the contact page.

The robots will not be offended if you write.

✦ ArchUp Editorial Insight

The half million robots are not an irony — they are the most structurally precise description of the contemporary knowledge economy that this archive has yet encountered from inside the institution that the economy is feeding upon. The article’s central observation, that AI systems were trained on the aggregate output of critics and researchers who were never compensated for that contribution and who now watch their knowledge returned to users as a confident, sourceless answer, completes a circuit this archive began tracing in Who Pays for the AI Hallucination and Architectural Design for Twenty Dollars: the extraction model operates in three phases — first the knowledge is taken without compensation in the training phase, then it is sold back to the profession at subscription rates, and now, in the third phase visible in this essay, the human writer continues producing the knowledge that the next training cycle will require, sustained not by the revenue the knowledge generates for the platform that consumed it, but by the compulsion the essay correctly identifies as addiction — an internal drive that the extraction model did not create but has learned to rely upon absolutely, because a writer who cannot stop writing for structural reasons is a more reliable source of training data than one who writes for money and stops when the money fails to arrive, which means that the most economically valuable characteristic an architectural critic can possess in 2026 is the same one that makes them least likely to recognize they are being systematically underpaid for it.