(in which the maze is named, but no longer possessed)

In an office lit by fluorescent panels, a form is completed and placed in a tray. The form is correct, as far as its author can tell. The handwriting is legible. Every required box has been filled. And yet, weeks later, it returns with a mark in red: resubmit. No explanation accompanies it. Only a reference number, and a note indicating that “the system” could not process the request.

By the early years of the twenty-first century, this experience had become familiar. The system was everywhere, invoked constantly, rarely seen. It determined access, eligibility, sequence. It did not argue or persuade; it accepted or rejected. Its decisions appeared neutral, procedural, final. One did not confront it directly. One navigated around it.

Long before personal computers entered the home, computing had already reorganised everyday life. From the 1960s onward, large institutions—governments, corporations, universities—began to rely on mainframe computers to process information at scales previously unimaginable. Building on principles formalised by John von Neumann, whose work helped define the modern computer as a general-purpose system, these machines occupied entire rooms, required specialised operators, and spoke languages few people understood. For most citizens, they remained invisible. Their presence was felt only through their effects.¹

To interact with these systems was not to use a machine in any ordinary sense. It was to submit oneself to a process. Information moved through predefined channels, advancing or halting according to rules embedded deep within the system’s logic. Each step depended on the successful completion of the last. Errors did not announce themselves clearly, they surfaced only as delays, denials, or requests to begin again.

The experience was curiously labyrinthine. One did not walk corridors, but one advanced through stages. Each form, queue, or approval acted as a junction. Progress was possible, but never guaranteed. There were no monsters, no dramatic obstacles—only procedure and logic. Here, language no longer points outward, but inward—each term leading only to another field, another condition, another required entry. The maze was not built to trap, but to sort.

Bureaucracy had always functioned this way to some extent. Max Weber described it as a rational system designed to eliminate arbitrariness through rules and hierarchy.² But when bureaucracy met computing, something changed. Rules could now be enforced automatically, without discretion. Pathways could be closed invisibly. The maze acquired walls that adjusted themselves.

In these systems, knowledge did not equal power but instead compliance did. Knowing where one was rarely helped, what mattered was whether the system recognised the position. Appeals existed, but they often looped back into the same process. One could follow the rules perfectly and still fail to advance, never quite learning why.

What distinguished these computational labyrinths from their architectural counterparts was their lack of spatial presence. There was no building to point to, no map to consult. The maze existed as logic, distributed across terminals, cables, and protocols. Its boundaries were difficult to perceive because they were not marked by walls, but by permissions.

The philosopher Michel Foucault once argued that modern power operates less through spectacle than through administration—through the quiet organisation of space, time, and behaviour.³ Computing amplified this tendency. Decisions once made by clerks were now made by systems. Responsibility dispersed. No single actor could be addressed.

In this environment, the phrase “the computer says no” entered common usage, half-joking, half-resigned.⁴ It captured a new kind of authority: impersonal, precise, and oddly unassailable. The system did not punish, it simply failed to respond.

For those who worked inside these institutions, the experience was no less maze-like. Programmers, analysts, and operators understood fragments of the system, rarely the whole. Code accumulated over decades. Documentation lagged behind implementation. Entire processes persisted because no one could confidently dismantle them. The labyrinth grew not by design, but by accretion.⁵

Users learned strategies. They copied forms. They memorised reference numbers. They developed informal guides to navigating official channels. Success depended less on understanding the system than on learning how to move through it without attracting error.

This navigation was temporal rather than spatial. Progress took time, sometimes a great deal of it. Waiting became part of the process. The maze extended into calendars and inboxes, into the gaps between responses. One advanced not by walking, but by enduring.

By the 1990s, personal computing had begun to transform everyday life. Graphical interfaces promised clarity. Icons replaced commands. Machines became friendly. Yet this friendliness masked a deeper continuity. While individuals gained access to personal devices, the systems that governed employment, finance, education, and security remained centralised and opaque. The mainframe did not disappear, it receded.

What changed was perception. The maze became harder to recognise because it no longer looked like one. Interfaces suggested immediacy, even as underlying processes remained complex and conditional. The system appeared closer, but understanding did not necessarily follow.

Around this time, theorists of cybernetics and control began to describe societies less in terms of institutions and more in terms of systems—networks of feedback, regulation, and response.⁶ Gilles Deleuze suggested that disciplinary structures were giving way to “societies of control,”⁷ where boundaries were flexible but continuous, and modulation replaced enclosure.

The mainframe labyrinth exemplified this shift. It did not confine bodies to spaces, it guided them through procedures. It did not demand obedience, it required compatibility. Those who could not be processed simply remained where they were.

After the turn of the century, these systems expanded rapidly. Security checks multiplied. Databases interlinked. Identity itself became a navigable structure, assembled from records and credentials.⁸ To move through the world increasingly meant moving through systems that recognised, verified, and logged.

Yet despite their reach, these labyrinths rarely became objects of reflection. They were experienced piecemeal, as inconveniences or frustrations. Their overall shape remained abstract. One encountered only the next step, never the design.

This invisibility was their greatest strength. Unlike walls or gates, procedures do not provoke resistance. They appear neutral, even inevitable. One does not rebel against a queue, one waits.

It is tempting to see these systems as failures of design—too complex, too rigid, too slow. But this misses the point. Their purpose was not elegance, but reliability. They were built to persist, not to be understood. The maze was not an error, it was a feature.

Standing at a service counter, watching a clerk type information into a terminal, it is possible to glimpse the system briefly—not directly, but in reflection. The clerk waits. The screen refreshes. Something happens elsewhere. The result appears.

In that moment, the labyrinth reveals itself not as a place one enters, but as a condition one inhabits. A structure that does not require walls because it is made of sequence. A maze that does not need to be seen in order to be followed.

Eventually, these systems would become more personal, more portable, more responsive. But the logic they established would remain. The corridors would multiply. The interfaces would change. And the act of navigation—procedural, patient, unresolved—would continue.

The maze no longer required architecture. It had learned how to operate on its own.

---

1. On the early development of mainframe computing and its institutional role, see Paul E. Ceruzzi, A History of Modern Computing (1998), which traces the shift from room-sized machines to distributed systems. ↩︎
2. Max Weber’s analysis of bureaucracy as a rational, rule-bound form of organisation appears in Economy and Society (published posthumously in 1922), and remains foundational for understanding administrative systems. ↩︎
3. Michel Foucault’s account of administrative power and spatial organisation is developed in Discipline and Punish (1975), particularly in his analysis of how modern institutions regulate behaviour through structure rather than force. ↩︎
4. The phrase “the computer says no” gained popular currency in the late twentieth century as a shorthand for automated decision-making, reflecting public ambivalence toward impersonal systems of authority. ↩︎
5. On the accumulation and persistence of legacy code within large systems, see Ellen Ullman, Close to the Machine (1997), a memoir that captures the lived experience of working inside complex computational environments. ↩︎
6. Norbert Wiener’s Cybernetics (1948) laid early groundwork for thinking about systems of control and feedback, influencing later social and technological theory. ↩︎
7. Gilles Deleuze introduces the concept of “societies of control” in his short essay “Postscript on the Societies of Control” (1992), arguing that modulation replaces enclosure as the dominant mode of regulation. ↩︎
8. On the growing role of databases, identity systems, and administrative computing in late twentieth-century life, see David Lyon, Surveillance Society (2001). ↩︎