AI : Summary
This session combined three intertwined demonstrations with a wide-ranging discussion on spatializing knowledge. First-time participant Benja demonstrated a prototype writing tool that visualizes deeply nested parenthetical clauses as stacked layers, addressing working-memory limits in complex prose. This was followed by a Game of Life knowledge-sculpture prototype for visionOS and a demonstration of a live Earth-and-space environment featuring real-time satellites, extrudable terrain, and the ability to walk inside the planet’s layers. The group then turned these demos toward a concrete challenge: designing a compelling spatial reading experience for the roughly thirty papers of the upcoming ACMHypertext Conference proceedings, which broadened into questions of annotation, curation, naming, and what a community’s unit of published conversation should be.
AI: Highlights & Insights
Center-embedded clauses expose working memory as a core constraint on processing complex text: two levels of embedding are readable, three collapse into incomprehensibility. A tool that assigns each clause its own visual layer lets a reader retain the top-level sentence while diving arbitrarily deep, effectively converting a linear string into a navigable hierarchy.
Writing and reading were reframed as fundamentally different modes of text navigation: reading is passive following, while writing is projective, imagining the clause being produced. Tools for thought may serve writing that is never intended to be read at all.
The nested-clause approach was connected to stretch text, Ted Nelson‘s trails, Vannevar Bush, telescoping text, visual-syntactic text formatting, and zooming user interfaces, with the observation that its novelty lies in the simplicity of authoring — plain brackets — rather than requiring a dedicated structuring tool.
The metaphor of “pitons” emerged: clipping anchor points to the wall of a rabbit hole while descending a digression, so one can close parentheticals and zip back to the original train of thought at any level.
On 3D knowledge sculptures, a design tension surfaced between connection lines as clutter versus lines as necessary orientation when space lacks a clear reference grid. Proposed resolutions included fading or toggleable lines, spatial-memory-based layout consistency, and gaze-activated highlight regions that remain invisible until looked at.
A significant realization about mixed reality: walking inside a virtual Earth in one’s own room via passthrough delivers the immersive benefit of VR without the danger of collision — something described as unimaginable two years earlier.
The claim that AI-assisted coding has collapsed the cost of prototyping to roughly the cost of a live demonstration was framed as transformative for the community’s entire intellectual life, with the phrase “fluidity of ability” coined to describe moving effortlessly between languages, platforms, and hardware.
A counterpoint tempered the enthusiasm: geographic visualization rests on hard data with an inherent coordinate system, whereas knowledge has no such given structure. Moving from describing an existing landscape to creating an unmapped one is a categorical leap, and any shared knowledge space will face disagreement about what to prioritize and connect.
A tension emerged between provocation and practicality — “from wow to work.” One position held that spectacle inspires competition and adoption; the other, illustrated by a vending-machine anecdote, argued that dramatic experiences designed for first encounters exhaust users who must repeat them dozens of times, and that good design (as in hand-gesture menus) can be simultaneously marvelous and fast.
Level-of-detail summarization was proposed as the key navigation device for a spatial proceedings reader: from a distance, thirty papers cluster into a few semantic groups; approaching reveals titles, then abstracts, then full text — a semantic zoom for literature.
What a reader takes out of a paper (quotes, annotations) and gives to a paper (judgment, elevation) were recognized as nearly functionally identical knowledge objects, suggesting annotations should be treated as first-class objects that remember their exact provenance, enabling ad hoc views organized by author, quote, or disagreement.
A distinction was drawn between annotation as embedded marking within a document and annotation as separate writing about a document — transclusion versus linked commentary — reflecting a computer-science versus humanities divide in conceptualizing the act.
The notion of “binding” was invoked as richer than linking: people become bound through shared places, meals, and papers, suggesting conference conversations could be bound to the papers that occasioned them.
Spatial memory was linked to abstract memory in the brain, with the further observation that altruism and future-oriented thinking share neural machinery — offered as connectome-level support for why spatial knowledge tools and generous communities may go together.
AI summarization was identified as a threat to attribution and linear intellectual history: summaries strip provenance, forcing readers to reconstruct sources through keyword search. The proposed remedy is AI summaries that maintain transclusion or hyperlinks back to authoritative origins.
Conferences were critiqued as poorly designed for knowledge transfer: the most important conversations happen in hallways, recordings are too slow to mine, and what is wanted is a map of a conference’s ideas and their connections to one’s own work, so attention can be allocated deliberately.
On naming and framing, the group recognized that frames both enable and restrict: a journal is a deliberative, curated funnel, while these meetings are free association, so treating a transcript as a journal issue imports assumptions that may not fit. The solid–liquid–aerial network metaphor captured the balance sought: too solid never changes, too aerial never gains purchase, liquid carves new channels while retaining structure.
The Total Perspective Vortex was offered as a metaphor for total information exposure as torture, framing the community’s central mission as curation — gaining mastery over overwhelming abundance rather than merely accessing it.
A question was raised about whether, if everyone consumes content through AI, we will end up writing for AI — answered with the suggestion that we will adapt our conceptual continuities rather than simply switch audiences.