Permanently Under Construction: Hypertextual Friction in an AI-Augmented Zettelkasten
TL;DR
It is suggested that hypertext’s role in scaffolding structured thinking is amplified rather than diminished by generative AI, provided the system demands co-construction rather than consumption.
Abstract
Contemporary AI tools for knowledge work encourage users to query and consume rather than construct and connect, risking a loss of the human agency that makes such work intellectually valuable. This paper presents an autoethnographic case study of an AI-augmented Zettelkasten, co-constructed within Obsidian using Claude Code. Through daily use over six months, the system and the researcher’s practices co-evolved via a bootstrapping process in which the system’s own conceptual resources were used to theorise its design, derive explicit values, and audit its own workflows. The resulting system operationalises hypertextual friction: deliberate points of interpretive effort where AI-generated proposals demand human curation before entering the knowledge network. The paper argues that these curatorial decisions are not obstacles to thinking but the site where thinking occurs, and that they are experienced as intrinsically rewarding rather than as cognitive cost. While the system works best for the person who built it (a builder’s advantage), this reflects an existentialist commitment that hypertext has always required of its users. These findings suggest that hypertext’s role in scaffolding structured thinking is amplified rather than diminished by generative AI, provided the system demands co-construction rather than consumption.
