The Cathedral That Refused to Become a Bazaar

Forty-five years on, Ted Nelson's 1981 report reads less like a technical proposal and more like a prophecy delivered in the wrong dialect — every major noun correct, every verb conjugated in a tense the future declined to use. Nelson foresaw that screens would become the primary surface for reading and writing. He foresaw that documents would link to other documents in webs of reference and response. He foresaw that publishing would become electronic, universal, and continuous. He foresaw that the struggle over information architecture would be a struggle over freedom itself. The dedication to Orwell was not decorative. What he could not foresee — or perhaps refused to accept — was that the world would get all of this in the ugliest, most expedient, most commercially captured way imaginable, and that most people would be fine with it. The Web as it arrived was not Xanadu. It was a parking lot built on the site where a cathedral had been planned, and the parking lot worked.

The specifics of Nelson's prescience are almost painful in their accuracy. The "universal electronic publishing system" is recognizable as the ambition that Tim Berners-Lee partially realized a decade later, minus everything Nelson cared about most: transclusion, bidirectional links, micropayment for reuse, and the preservation of a document's "true intrinsic structure." Nelson wanted every quotation to remain tethered to its source, every version tracked, every act of reading to be an act of navigating a living web of attribution. In 2026, we have instead a world where text is scraped by language models that dissolve attribution entirely, where links rot within months, where the dominant mode of electronic publishing is a social media post that exists in no archive and answers to no structure. The Arghive — Nelson's imagined universal repository — never materialized. What materialized was the Wayback Machine, a heroic afterthought, and a thousand paywalled silos. Nelson diagnosed the disease with perfect clarity. His prescription required a patient who did not exist: a civilization that valued coherence over convenience.

The blind spots are era-specific but revealing. Nelson's framework is relentlessly textual. Images, audio, and video are acknowledged but never centered; the assumption is that the intellectual revolution will be a revolution of writing and reading. There is no anticipation of a culture that would move from text to image to short-form video as its dominant information medium, nor of a public that would prefer watching to reading by a ratio that would have struck Nelson as civilizational collapse. His faith in the individual user's desire for control and structure now seems almost aristocratic — not wrong in principle, but wrong about the distribution of that desire across a population. He also assumes, with the optimism of the early personal computing movement, that better tools will produce better thinking. The last decade has been a controlled experiment in the opposite hypothesis. More interesting is what Nelson gets right about power. His insistence that information architecture is political, that the design of systems determines who can speak and who can be heard, anticipated the platform governance debates of the 2010s and 2020s with an exactness that most technology writers of his era could not approach. The Orwell dedication earns its place.

Nelson occupies a peculiar position: the prophet without a church. He gave hypertext its name and its most rigorous early theorization, and in doing so he shaped the vocabulary that Berners-Lee, Marc Andreessen, and every subsequent web architect inherited, even as they abandoned his architecture. Xanadu is the road not taken that haunts every conversation about what the internet should have been. Vannevar Bush's "As We May Think" is the ur-text; Nelson is the first full-length sermon. Everything that came after — from the Web to wikis to the current generation of tools-for-thought software — exists in dialogue with Nelson's vision, usually by simplifying it past recognition. The 1981 report is the moment the vision was most fully articulated and most politically charged, before decades of non-delivery turned Xanadu into a punchline. Reading it now, the punchline dissipates. The comedy was never in Nelson's ambition. It was in ours.

If Nelson was right that the structure of information systems determines the structure of thought, and if the systems we built were the ones that optimized for engagement, extraction, and ephemera rather than for connection, attribution, and depth — then what exactly have we been thinking with for the last thirty years, and would we recognize the difference if we stopped?