AI: Summary
This session explored the intersection of AI in education, the crisis of academic publishing and document ownership, and the question of what a future-proof knowledge document format might look like. The conversation moved from how AI levels the playing field for students and reshapes teaching toward mastery-based learning, through the social and technical failings of current web and publishing infrastructure — siloed libraries, broken hyperlinks, pay-to-publish incentives — and arrived at a substantive design discussion about whether ePub, a new format, or enriched HTML could serve as the foundation for ownable, citable, shareable units of knowledge in a post-PDF world.
The writing-as-thinking principle was reaffirmed through the experience of submitting an ACM Hypertext paper, with the observation that AI is effective at citation-checking and factual auditing but still falls short on creative and conceptual judgment — the human must still determine whether a source is actually relevant to the argument being made, not merely adjacent to it.
The idea that AI functions as an educational leveler provoked a significant reframing: historically, wealthier students could buy tutoring and essay-writing services, while now AI provides equivalent cognitive scaffolding to anyone. The implication is that institutional resistance to AI in education may partly reflect discomfort with the democratisation of advantages that were previously class-gated. This was paired with a shift toward mastery-based learning, where students must demonstrate genuine understanding rather than regurgitation — a pedagogy that becomes both more necessary and more achievable with AI support.
A guest participant from the Endangered Alphabets Project, Tim Brookes, raised the question of embodiment and manual skill: young people in economically depressed regions are rejecting trades in favour of screen-based learning, and the decline of fine motor skills represents a cultural and cognitive loss. The counter-observation was that spatial computing environments like Author and Knowledge Space may offer a synthesis — making intellectual work itself a manual, motor-cognitive act through spatial arrangement and embodied interaction, even within a digital medium.
A productive tension emerged between web-native and document-native worldviews. One participant argued that the web is the only platform that will ultimately matter for computing, and that many problems framed as technical are actually social and political — peer review corruption, citation gatekeeping, and publisher rent-seeking are not failures of HTML but of institutions. The counterpoint was that practical ownership demands a portable, self-contained artifact: something you can put on a hard drive or thumb stick, independent of any publisher, platform, or server. The web’s security model, which prevents web pages from accessing local file systems, was identified as a concrete barrier to building integrated knowledge environments on the open web.
The discussion of ePub as a potential bridge format surfaced genuine architectural interest. ePub 3 is structurally HTML, CSS, and XML metadata in a single distributable package — reflowable, parseable by AI, and carrying rich semantic markup. However, reader fragmentation is severe, JavaScript support is inconsistent to nonexistent across readers, and citation workflows are poorly supported. The suggestion of standoff markup — raw text separated from annotation layers, bundled inside a single package — was proposed as a cleaner architecture that could live inside an ePub-like container or a wholly new format.
The session’s sharpest unresolved question concerned the gap between what ACM and similar libraries claim to offer and what they actually deliver: a recent ACM article was shown to lack even a basic outline, offering only a bare PDF with no meaningful digital affordances despite being a digital-native publication. The observation that ACM is selling legitimacy rather than publication quality crystallised the broader problem — the credibility infrastructure of academic knowledge has drifted far from the technical infrastructure needed to make that knowledge findable, verifiable, and connectable.
The idea of citing ideas rather than whole documents was raised as a closing provocation — a shift that would require a fundamentally different citation architecture but could better reflect how knowledge actually works.
The practice of generating AI songs from meeting transcripts was discussed as a workflow innovation, with a detailed prompt structure shared for producing jazz-rap lyric summaries of conceptual sessions — a form of knowledge encoding that someone outside the community described as resembling the work of a real South African rapper-linguist, raising interesting questions about where AI-generated cultural forms begin to converge with human creative traditions.
Song
AI: Resources Mentioned
ACM Hypertext Conference — paper submission deadline referenced as the day of the meeting
Sonix.ai — used for video transcription (https://sonix.ai)
Suno — AI music generation platform used for creating session songs
Claude — AI assistant used for research, citation-checking, and paper co-writing
YouTube Inspirations tab — new AI-powered content suggestion feature noted on YouTube
NotebookLM — used for uploading open educational resources as a searchable student textbook replacement
SIFT Method — information literacy framework developed by Mike Caulfield (Stop, Investigate, Find, Trace)
How to Lie with Statistics — book from the 1950s, recommended as available on the Internet Archive
Endangered Alphabets Project — founded by Tim Brookes (https://www.endangeredalphabets.net)
Origin Private File System — web API for disc-like read/write operations, referenced in context of Universal Scene Description files (https://developer.mozilla.org/en-US/docs/Web/API/File_System_API/Origin_private_file_system)
SharePlay — Apple co-presence framework for shared visionOS experiences
iWeb — referenced as a historical example of accessible web page creation tools
Dave Winer and RSS — story of getting Mozilla to adopt RSS mentioned
Federated Wiki — link shared: http://petedaguru.dojo.fed.wiki/view/welcome-visitors/view/finding-truth-in-the-age-of-ai
The Holdovers — film with Paul Giamatti, recommended
For All Mankind — Apple TV+ series discussed
The Martian — film by Ridley Scott, recommended
Dan Jones — book on the Crusades mentioned as current reading
Ender’s Game — referenced in context of communication potential
Future Text Lab session page: https://futuretextlab.info/2026/04/27/27-april-26/
Recording of previous day’s event with Tim Brookes’s community: https://youtu.be/QWCflUoqNfA
Musical video “I Want My Flying Car”: https://www.youtube.com/watch?v=cLvy7sYOOiY
Additional video shared: https://youtu.be/vzm6pvHPSGo