Literary Machines
Source: https://archive.org/details/literarymachines0000nels ↗
Nelson's self-published, endlessly revised manifesto describes Project Xanadu — a hypertext system conceived in the 1960s that envisioned two-way links, version tracking, micropayments for authors, and transclusion as alternatives to copying.
The book is essential for understanding the road not taken: when Berners-Lee designed the World Wide Web a decade later, he deliberately chose one-way links and simplicity over Nelson's richer but more complex architecture.
Nelson's writing is passionate, eccentric, and sometimes maddening, but his critique of hierarchical file systems and his vision of interconnected documents were decades ahead of their time.
Many problems the web still struggles with — broken links, content ownership, attribution — are precisely the ones Xanadu was designed to solve.
Reading Nelson today is a reminder that the systems we use are not inevitable but the product of specific tradeoffs made by specific people.
Central argument
Nelson argues that documents should exist in a universal, interconnected repository where links are bidirectional, content can be transcluded rather than copied, authors are micropaid for reuse, and every version is preserved — a system he called Project Xanadu. His central thesis is that the hierarchical file system and the act of copying are architectural mistakes that destroy provenance, attribution, and accountability. The web we inherited is not the only possible web; it is the result of Berners-Lee choosing simplicity and adoption speed over the richer, more rigorous architecture Nelson had been developing since the 1960s.
Critique
The deepest tension in Nelson's work is that Xanadu's comprehensiveness is precisely what prevented it from existing: a system that solves attribution, versioning, broken links, and micropayments simultaneously requires global coordination and buy-in that no single actor can compel. Nelson's critique of simplicity as a tradeoff reads more convincingly in hindsight than it would have as a design brief, and a thoughtful reader might ask whether a system this dependent on universal adoption could ever have bootstrapped at all — making the web's 'inferior' architecture not an accident but a rational response to the cold-start problem.
Why it matters for product
For a CPO, Xanadu is a masterclass in the difference between architectural vision and adoptable product: Nelson was right about the problems and wrong about the tractability of solving them all at once, which is a failure mode product leaders reproduce whenever they design for the ideal steady state rather than for the adoption path. More concretely, the broken-link and attribution problems Nelson identified in 1981 resurface today in content platforms, internal knowledge systems, and AI training data — product teams still make the same tradeoff between simplicity of creation and integrity of connection, usually without acknowledging it as a choice.