AI: Summary
The session ranged across the lived experience of working with text and knowledge artifacts in XR, the challenge of representing time in spatial environments, the emerging gestural grammar of visionOS interactions, and the broader socioeconomic responsibilities that come with building new knowledge tools. Participants debated how color, scale, and depth might encode temporal information, explored the tension between computational precision and the necessary vagueness of human knowledge, and reflected on whether technology alone can deliver on its liberating promise — or whether intentional economic and cultural thinking must accompany the tooling.
AI: Main Topic
The primary discussion centered on how to bring text, annotations, and temporal metadata into XR in a way that respects the fluid, imprecise nature of human knowledge. Frode Hegland opened by demonstrating and narrating the current state of his Author / visionOS spatial-text prototype, with Jonathan Finn wearing an Apple Vision Pro in the room and reporting on the bodily experience of walking into information. The group examined new gestural interactions — look-and-pinch to select, double-pinch to open without selecting, pinch-in-pull-out to isolate connected items, and pinch-in to snap an object back to its origin plane — and debated how time should surface as a layout dimension versus a filtering or coloring mechanism.
AI: Highlights
Tom Haymes shared that Gemini built him a bespoke project-management framework synthesized from cybernetics, the Gutenberg Parenthesis, and historical wargaming — not drawn from any single book but custom-assembled from his conversational history. He described this as a sign that AI can create personalized cognitive tools rather than merely retrieve existing ones.
Karl Schroeder’s aphorism, relayed by Tom Haymes in the chat, crystallized a core tension: “A codex teaches the reader to think in places. An ebook teaches the reader to think in streams.”
Frode Hegland reported that Phil Gooch used Claude AI to build an entirely self-contained game engine in JavaScriptrendering on the web, then generated a multiple-choice adventure from a Shakespeare passage in roughly ten minutes — including the AI pushing back on narrative choices, suggesting it start at a different section of Hamlet. Frode framed this as evidence that AI hallucination is largely resolved for well-scoped research tasks.
Ken Perlin demonstrated a WebXR text experiment using signed distance fields, showing the first page of A Tale of Two Cities that fades gracefully with distance and becomes crisp as the viewer walks toward it. He assigned his VR class to design freely with scale-independent text in a headset.
Astral Druid described AI as potentially a “universal translator” across expertise domains, while Peter Dimitrios countered with “universal fragmentor” — a tension Ken Perlin addressed by pointing to his new blog post on the topic.
AI: Insights
Brandel Zachernuk argued that text’s superpower is its capacity for generality — the word “animal” functions perfectly even though no visual image of a generic animal exists — and that spatial computing’s drive toward specificity (coordinates, pixels, precise layouts) actively works against this strength. He urged the group to “fight for” representational vagueness inside a medium that structurally demands precision.
Mark Anderson’s counsel on “soft coupling” for time data emerged as a design principle with broad implications. His argument was that deterministic, fixed-field approaches to time inevitably trample the very information you are trying to surface, because real-world temporality arrives in overlapping, ambiguous, multi-series strands. Binning — choosing eras rather than linear scales — and keeping coupling loose were offered as hard-won lessons from two decades of practice.
Jonathan Finn, speaking from inside the headset, articulated an experiential paradox: walking toward information feels more natural than pulling it toward you, yet the physical cost of walking is high. He proposed that the system might accelerate the approach — meeting the user halfway — blurring the line between gesture and locomotion. Peter Wasilko extended this into the idea of “active” data objects that sense proximity and flock toward the user, echoing force-directed graph behaviors.
Ken Perlin reframed the group’s purpose by insisting that the community should not merely build tools but explicitly think about economic models — asking who benefits and how distributed, small-business-scale production might be enabled through spatial AI, rather than defaulting to platform monopolies.
Brandel Zachernuk pushed back on Tom Haymes’s claim that coding was the main barrier to personal universe-building, arguing instead that the real constraint is the “deliberative space” — the time and cognitive room people have to engage with powerful tools. Phones are adequate for small tasks but hostile to deep work; desktops offer capability but not invitation. He expressed hope that VR might uniquely press on both the spatial and the temporal dimensions of this problem simultaneously.
Mark Anderson drew an analogy from crisis-center teamwork — military, medical — where trust enables people to speak across knowledge boundaries rather than defending disciplinary turf. He connected this to the observation that transdisciplinary collaboration, not mere multidisciplinary co-presence, is what produces breakthroughs, and that the hypertext community’s reunification of its artistic and engineering wings consistently generates productive friction.
Frode Hegland proposed the Victoria and Albert Museum as a historical model for what a 21st-century XR knowledge commons might look like — a space originally built for working people, with gas lighting so they could attend after work, devoted not to aristocratic taste but to practical mechanical arts and inspiration. Jonathan Finn quipped that it was “Victorian VR” — casting a replica of Trajan’s Column to transport visitors to Rome.
The distinction between context and content sharpened during the session. Frode noted that when pasting defined concepts from one Author document to another, the receiving document should tag them as “context” so they can be hidden or revealed independently of new material — a small metadata decision with large implications for how people manage cognitive load in spatial environments.
Astral Druid introduced David Graeber’s historical lens: that ‘locking’ technologies roughly 5,000 years ago inaugurated a zero-sum competitive paradigm, and that the current moment may allow a return to more fluid, commons-based arrangements. He argued for “accessibility” over “affordability” and positioned libraries as existing commons infrastructure suited to hosting emerging spatial technologies.
AI: Resources Mentioned
Factify — smart document platform aiming to replace PDFs and DOCX, shared by Tom Haymes: https://venturebeat.com/infrastructure/factify-wants-to-move-past-pdfs-and-docx-by-giving-digital-documents-their
Cartographies of Time — book mentioned by Brandel Zachernuk and Mark Anderson, currently out of print, with one author at the University of Oregon
The Book of Trees: Visualizing Branches of Knowledge — mentioned by Peter Wasilko, who was considering an interlibrary loan
Timeline One: A History of Editing — book recently acquired by Brandel Zachernuk
Inventing the Future — book by the same author as Timeline One, about Apple‘s Advanced Technology Group, mentioned by Brandel Zachernuk
Almost Perfect — book by Pete Peterson about WordPerfect, recently finished by Brandel Zachernuk
Orality and Literacy by Walter Ong — mentioned by Frode Hegland
The Dawn of Everything and Debt: The First 5,000 Years by David Graeber — mentioned by Astral Druid
Fall by Neal Stephenson — mentioned by Tom Haymes in the chat
The Last Economy — free book by one of the people behind Stable Diffusion, shared by Tom Haymes: https://ii.inc/web/the-last-economy
Discovering Digital Humanity — Tom Haymes’s own work, referenced in the chat
Ken Perlin’s blog post on the “universal fragmentor” theme: https://blog.kenperlin.com/?p=28201
Apple 1984 Super Bowl ad directed by Ridley Scott — discussed by Ken Perlin
Apple Knowledge Navigator concept video (1987) shared by Frode Hegland: https://youtu.be/ecP_xoKrQ9Y?si=thpi7owVnML-yCAH
Karl Schroeder — quoted by Tom Haymes: “A codex teaches the reader to think in places. An ebook teaches the reader to think in streams.”
Douglas Adams — quoted by Tom Haymes: “Time is an illusion, especially lunchtime.”
Jaron Lanier — referenced by Ken Perlin for his vision of AI as interface between human expertise domains
Doug Engelbart — referenced by Frode Hegland as foundational influence
Alan Kay — referenced by Frode Hegland (“the best way to predict the future is to invent it”) and Tom Haymes (monkeys and telescopes anecdote)
Phil Gooch — formerly of Scholar, mentioned by Frode Hegland for building a Claude-powered game engine
Dave Millard — referenced by Frode Hegland as a knowledgeable colleague
The Victoria and Albert Museum, London — discussed by Frode Hegland and Jonathan Finn
The Metropolitan Museum of Art, New York — mentioned by Brandel Zachernuk regarding digitized Asian scrolls
Meta Quest 3 — Ken Perlin purchased 16 units for his students at roughly $300 each
Apple Vision Pro (M5) — discussed by Frode Hegland, Brandel Zachernuk, and others
Song
This track is an AI orchestrated piece inspired by the transcript of this meeting, meant as a fun provocation to further thought. (suno.com)