For today’s meeting: If we have free-flowing text documents in XR (not PDF) and the documents are self contained (not HTML), with useable visual and semantic markup (ePub maybe), what views and interactions might be useful? As a reminder of my mentors: Doug’s work, to my memory, was mostly about addressing and jumping. He used visual hierarchy. Ted Nelson’s work on hypertext was visual with parallel documents visibly connected. This makes me think about how, if we abandon PDF for reading and focus on text, we can have a foldable structure with connections coming out. We might want to think of an ‘origami’ aspect to the text.
AI: Summary
This session ranged across the conceptual and practical challenges of making text spatial, from live demos of the Author app running on Apple Vision Pro to a sustained debate about what document format — HTML, Epub, JSON, or something new — could serve as an open interchange standard for XR-based knowledge work. The discussion was anchored by Frode Hegland’s demonstrations and framing, but branched into deep territory on foldable document metaphors, the philosophical gap between available tools and human willingness to use them, the tension between long-term standards work and short-term experimentation, and nostalgia for abandoned paradigms like OpenDoc. The meeting also touched on upcoming community events, a Wikipedia-based XR demo by Ken Perlin, and Brandel Zachernuk’s perspective on the years-long timeline required to make the spatial web a genuine standard.
AI: Main Topic
The primary topic was authoring and interacting with documents in spatial environments, explored through the lens of what the community can build now versus what must wait for formal standards. Frode Hegland opened by presenting a demo of his Author software running on Vision Pro, showing how meeting-record nodes (index cards representing people, topics, and terms) can be selected, sorted, focused, hidden, liked, and snapped back to their original positions in 3D space. He framed this by revisiting the foundational visions of Doug Engelbart (flexible views, hierarchical navigation, the 1968 NLS demo) and Ted Nelson (parallel documents, visible connections), asking the group to imagine a new free-flowing document format — richly marked up, citable, and secure — that could support these kinds of spatial interactions. This led to a proposal he called “origami text”: the idea that a document should be foldable in multiple ways (outlines, glossaries, references, timelines, citation layers), where different views are different folds of the same underlying material. The second half of the meeting shifted to the practical question of what format could carry this spatial metadata between environments. Mark Anderson made a detailed case for clean, minimal HTML over Epub, and Brandel Zachernuk agreed on the representational level while cautioning that the standards work required to make spatial HTML real is a multi-year undertaking involving W3C negotiations. The session ended with Mark Anderson and Frode Hegland converging on the idea that JSON embedded within HTML could serve as a near-term interchange format for spatial node positioning.
AI: Highlights
Frode Hegland announced that Future of Text and the ACM Hypertext Conference (HT’26) will both take place in London in September 2026, sharing a building but not formally co-located. He is in discussions with the organizers about building an XR representation of the conference — a spatial timeline corridor where papers and events are listed and interactable — and proposed this as the community’s main topic for March.
Ken Perlin demonstrated a deceptively simple WebXR tool built for his VR class at NYU that pulls Wikipedia articles via their API and constructs a navigable tree of links. He emphasized that the underlying code is “embarrassingly simple” and that Wikipedia’s plain-text heading markers (equal signs) already provide the semantic structure needed for XR views. The demo URL is cs.nyu.edu/perlin/wikipedia.
Frode Hegland revealed that Claude spontaneously generated an interactive application illustrating the concepts of Engelbart’s view specs and Nelson’s parallel pages when he described the community’s work — without being asked to build an app. He also used Claude to identify a contact at MIT, which turned out to be an old friend, calling it a “crazy world.”
Brandel Zachernuk shared a VR representation of the Young Chang Heavy Industries Lotus Blossom (a Korean artist duo’s work), available at zachernuk.neocities.org/2026/yhchang/, and noted it runs on Quest and other headsets. He invited Ken Perlin to use it in his VR text discussions.
Tom Haymes shared a Jerusalem Post article about a startup attempting to make PDFs intelligent through embedded code, which recently IPO’d. Frode Hegland was skeptical, noting that security restrictions in reading software would likely prevent meaningful progress, drawing on his own PhD experience of trying to extend PDF functionality.
Frode Hegland mentioned he recently spoke with Bruce Horn, an early Apple employee, encouraging him to consider building a spatial Finder for VisionOS, since the current desktop environment on Vision Pro is, in Frode’s words, “awful.”
Tom Haymes described how Gemini built him a personalized orbital task-management system designed specifically for his cognitive style — interests in gaming, systems thinking, and visual thinking — rather than a generic productivity framework. He said this experience made him realize that AI combined with spatial interfaces could create genuinely customized work environments, moving beyond the “lowest common denominator” of self-help systems.
Mark Anderson shared and demonstrated two versions of his ACM paper manually ported from PDF to the ACM HTML format — one with full CSS/JS and one stripped bare — to show that well-formed HTML degrades gracefully and remains perfectly readable without styling. The URLs were shared in the chat for the group to examine.
AI: Insights
A significant conceptual tension emerged between Brandel Zachernuk’s long-horizon view and Frode Hegland’s desire for near-term experimentation. Brandel was explicit that the question of how HTML becomes spatial is a W3C-level undertaking that “will be many years away, and it will be the foundation of the way that the web is brokered and presented for the next 50 years.” Frode pushed back not by disagreeing but by arguing that the speed at which major companies will colonize XR once it takes off — “an absolute meteor hitting the planet” — makes it urgent to establish open interchange now, even in primitive form. This tension was not resolved but may be the most productive friction point in the community: Brandel’s caution protects against premature ossification, while Frode’s urgency ensures something concrete gets built and tested.
Tess Rafferty introduced what may be the session’s most underexplored philosophical question: why do people not use the organizational tools already at their disposal? She framed it in terms of tactility and physicality — “this intrigues me because it’s stuff I can touch” — wondering whether the appeal of spatial interfaces is precisely that they restore a physical relationship to information that flat software has removed. Tom Haymes connected this to the deeper problem that sorting and organizing often takes more time than collecting, which creates a natural decay in organizational habits. This line of thinking suggests that spatial computing’s value may lie less in adding new capabilities and more in reducing the cognitive overhead of existing ones.
Mark Anderson’s insistence on the concept of “scope” for spatial views — that the useful working set is “somewhere south of about 100 notes” — and that beyond that threshold you are doing information visualization rather than hands-on manipulation, represents a practical design constraint the community will need to internalize. Frode initially seemed to push back, citing the additional controls (hide, sort, timeline) that expand capacity, but Mark clarified that they were actually in agreement: the point is not that you cannot have more data, but that the mode of interaction fundamentally changes at scale.
The stretch text discussion revealed an interesting disciplinary boundary. Mark Anderson was firm that stretch text, as defined by Ted Nelson in 1967, has a precise meaning — text that expands and contracts inline, requiring careful management of anchor points and surrounding prose — and should not be conflated with other forms of expandable content like endnotes or collapsible sections. Brandel Zachernuk pointed out that Wikipedia mobile already uses expanding and collapsing sections and “nobody really complains about it,” suggesting that the practical implementation may have outrun the theoretical framework. Mark acknowledged this but maintained the distinction matters for conceptual clarity, especially in a community that takes Nelson’s ideas seriously.
Brandel Zachernuk’s observation that “all text is hypertext” — attributed to his daughter via Terry Pratchett’s densely referential writing — reframes the entire spatial-text project. If text is already implicitly hyperlinked through allusion, citation, and reference, then the community’s work is not about adding hypertext to text but about making existing implicit connections navigable and visible. This aligns with Frode’s origami metaphor: the folds are already in the document; the technology needs to let you unfold them.
The game engine debate between Astral_Druid and Brandel Zachernuk exposed a fundamental architectural question for spatial computing. Astral_Druid proposed Godot as a base because game engines handle spatiality and interconnectedness natively. Brandel countered that any engine operating in direct rendering mode has total knowledge of user behavior, which creates an impossible trust boundary for multi-application environments. His analogy was clear: just as macOS mediates between applications, a spatial web needs a mediating layer that no single engine controls. This is not a disagreement about tools but about whether the spatial future looks more like a game (one engine, total control) or like the web (multiple agents, mediated trust).
The convergence on JSON-embedded-in-HTML as a pragmatic near-term format emerged organically from the discussion. Mark Anderson’s key insight was that well-formed HTML already has a DOM with structure, so unlike PDF, it does not need its structure restated externally. The spatial positioning data (x, y, z coordinates, IDs, metadata) can ride as embedded JSON within the HTML document, keeping everything in one portable file. This approach sidesteps the multi-year W3C process Brandel described while remaining compatible with it — a rare case where the pragmatic and principled paths do not conflict.
AI: Resources Mentioned
Ken Perlin’s Wikipedia XR demo — https://cs.nyu.edu/perlin/wikipedia (shared by Ken Perlin)
Brandel Zachernuk’s Young Chang Heavy Industries Lotus Blossom VR piece — https://zachernuk.neocities.org/2026/yhchang/ (shared by Brandel Zachernuk)
WebSpatial SDK and documentation — https://webspatial.dev and https://github.com/webspatial/webspatial-sdk (shared by Peter Dimitrios and Brandel Zachernuk, who noted the “unified rendering” concept page at https://webspatial.dev/docs/core-concepts/shared-space-and-spatial-apps)
Jerusalem Post article on making PDFs intelligent — https://www.jpost.com/business-and-innovation/article-884932 (shared by Tom Haymes)
World XO organization — https://worldxo.org/ (shared by Tess Rafferty)
Mark Anderson’s ACM paper in HTML format — https://dl.acm.org/doi/fullHtml/10.1145/3648188.3678215 (shared by Mark Anderson)
Mark Anderson’s manually ported paper (with ACM CSS/JS) — https://www.shoantel.com/proj/papers/acm-html/WhitherSH.html (shared by Mark Anderson)
Mark Anderson’s paper without CSS/JS — https://www.shoantel.com/proj/papers/WhitherSH.html and https://www.shoantel.com/proj/papers/acm-html/WhitherSH-nocssjs.html (shared by Mark Anderson)
Apple HotSauce / Project X — https://www.macintoshrepository.org/866-apple-s-hotsauce-technology-project-x- (shared by Peter Wasilko)
Cyberdog introduction article (archived) — https://web.archive.org/web/20210325144546/http://preserve.mactech.com/articles/mactech/Vol.12/12.02/CyberdogIntro/index.html (shared by Peter Wasilko)
Cyberdog team at Apple (archived) — https://web.archive.org/web/19961101214306/http://www.cyberdog.apple.com/group.html (shared by Peter Wasilko)
Cyberdog overview (archived) — https://web.archive.org/web/20050310094909/http://www.cyberdog.org/about/cyberdog/cyberway.html (shared by Peter Wasilko)
Snow Crash Chapter 5 (annotated) — https://genius.com/Neal-stephenson-snow-crash-chapter-five-annotated (shared by Peter Wasilko)
Ted Nelson’s stretch text original paper — from https://archive.org/details/SelectedPapers1977, page 27 (shared by Mark Anderson)
MRjs spatial web components — https://github.com/Volumetrics-io/mrjs (shared by Peter Dimitrios)
Mermaid-3D visualization tool — https://github.com/sunnydark/mermaid-3d (shared by Peter Wasilko)
Docling document conversion tool — https://www.docling.ai/ (shared by Peter Dimitrios)
Mozilla HTML cite element reference — https://developer.mozilla.org/en-US/docs/Web/HTML/Reference/Elements/cite (shared by Frode Hegland)
Knowledge LaTeX package documentation — https://us.mirrors.cicku.me/ctan/macros/latex/contrib/knowledge/knowledge.pdf (shared by Peter Wasilko)
Doug Engelbart — computing pioneer, NLS creator, 1968 demo (discussed extensively by Frode Hegland)
Ted Nelson — hypertext pioneer, stretch text inventor, Xanadu visionary (discussed by Frode Hegland, Mark Anderson, and Brandel Zachernuk)
Bruce Horn — early Apple employee, mentioned by Frode Hegland as someone he encouraged to build a spatial Finder for VisionOS
Ted Göransson — collaborated with Mark Anderson on stretch text implementation around 2010
Alan Kay — referenced by Tom Haymes in relation to Xerox PARC work where there were no apps, only tasks
Neal Stephenson — author of Snow Crash, referenced by Peter Wasilko
Terry Pratchett — referenced by Brandel Zachernuk (via his daughter) as an example of text that is inherently hypertextual
Song
For Website: This track is an AI orchestrated piece inspired by the transcript of this meeting, meant as a fun provocation to further thought.