Elements, Connections
AI: Summary
This session explored what the fundamental elements of a spatial knowledge environment should be, launched by an AI-assisted inquiry into neuroscience and psychology which concluded that connections, not objects, are the primary units of thought. The conversation wove through the reconstructive nature of memory, the failure of education systems to teach thinking rather than conformity, a sustained debate over whether original ideas exist, the problem of provenance and intellectual property for transclusion, and how documents might be restructured, connected, and lived in within AR environments such as visionOS.
Eric’s resource: https://github.com/UDerangell/EricsGarage/issues/2
The etymology of “node” as a knot on a string produced a reframing: the thread is the vital element and the knot merely sits upon it, implying that in any knowledge graph, nodes and connections should be interchangeable — a link displayable as a node and vice versa.
The primitive act of mind was proposed to be distinguishing — asserting that this differs from, or resembles, that — rather than asserting existence, aligning spatial knowledge tools with comparison and relation rather than object storage.
Because human memory is reconstructive rather than a retrieval of solid factual objects, a spatial system should not mimic storage of things; yet abstract concepts still require a graspable shape so a human hand can touch, move, and further define them in space.
A notable design principle emerged from AI dialogue: a system should ideally not permit adding an item without connections, making relation a condition of entry rather than an afterthought.
Making an AI argue against itself across separate instances — where one instance praised a claim and the next dismantled it — was surfaced as a genuinely fruitful method and framed as part of what literacy now means: learning to deal with other minds, including artificial ones.
A tension animated much of the session: the claim that there are no original ideas, supported by a seed metaphor in which every idea carries the genetic footprint of prior ideas and must be planted in soil made of other people’s thinking, against the view that denying originality demeans human imagination. A third position was proposed — neither agreeing nor disagreeing but “seeing it differently” — with the observation that no word exists for this stance.
A complementary argument located meaning not in discrete bits but in interactivity: memories, meanings, and ideas share an ontological interdependence and are reconfigured with every interaction, meaning emergence cannot be predicted from the pieces alone. The digital medium’s promise is that its morphologies of information can be more interactive than grey matter allows.
Independent parallel discovery — illustrated by a student reinventing reference-counting garbage collection without prior exposure — was offered as evidence that patent absolutism is unfair, and that experiencing replication through programming shifts people from patent to copyright preference.
Provenance of ideas was flagged as important for future humans, alongside the observation that intellectual property barriers make transclusion nearly impossible in practice, with historical friction over trademarked hypertext concepts cited as a lived example.
Education was diagnosed as an industrial system that rewards conformity: students learn to ask “what do you want me to think,” grades punish genuine dissent, and the hope that AI would supply the missing Socratic conversation in asynchronous learning proved harder than expected — the barrier is training and risk-aversion, not technology.
Augmentation was given a strict criterion: it only counts if the person grows from the process. Retrieving answers and forgetting them leaves you unaugmented; genuine augmentation shifts how you think and connect, pushing people up the taxonomy of thinking toward creation.
The recurring failure of hypertext history was restated: no good connection machine has ever been built because links remain manual and filtered through the blinders of the author, while teaching transmits objects rather than the connections between them. Relationships, it was argued, need to be first-class objects.
Against the proposal of personal glossaries transcluded and mapped between individuals, a counter-vision emerged of knowledge as directions rather than objects — magnets, push and pull — since some understanding, such as an hour-long emotional transmission of lived identity fear, may resist mapping entirely and require story, film, or multimodal AR.
Documents in AR were shown exploding into spatial sections, with marked sentences serving as the “first semantically valid unit” of each paragraph — a semi-structured evolution of NLS viewspecs — accompanied by agreement that the system must know what its parts are before views can be intelligently changed.
The phrase “knowledge sculptures” was embraced as romantic terminology for spatial thinking artifacts, alongside a catalogue of the costs of going “between”: between linked pages, media, minds, cultures, paradigms, and histories.
Edit decision lists from film were extended into a vision of remixable, transcludable media — including variable versions of the same shot to make media instantly rewatchable — and voice-first journaling was noted as a distinct brain modality whose ease of capture changes what gets recorded.
Two AI aphorisms were highlighted as memorable: that a foundational augmentation pioneer “built a cathedral and required you to worship in it,” and that humans “look at the world through a straw” — a reminder that human perceptual limits must anchor interface design.