July was the month the Lab’s centre of gravity shifted from talking about building to building while talking. Across the four sessions — 6, 13, 21 and 27 July — a single arc is visible: the cost of turning an idea into working software collapsed, and the conversation reorganised itself around what that makes possible and what it demands philosophically.
From specification to demonstration
The hinge was the validation of Claude working inside Xcode as a trusted part of the development workflow, demonstrated live to the Lab in mid-July. Where earlier sessions produced specifications to be handed to a programmer, July’s sessions increasingly ended with software built during or between the calls. The 13 July session made the framing explicit: AI as co-creator within the coding environment rather than a standalone generator — working the material like wet clay — and the 21 July session drew the conclusion out loud: tools are now cheap to make, the AI functions as a genuine partner in the environment where the work happens, and there is therefore no excuse not to build, test, and discover what an idea is actually worth. The 27 July session opened by enacting the claim, with a live walkthrough of building, running and notarising a native application by conversation alone, followed by a demonstration of a globe and solar-system application on Vision Pro.
What the sessions were actually about
The building was never the point in itself; it served a sustained inquiry into what a personal knowledge space should contain and how its contents should behave. The 6 July session set the conceptual terms: connections mattering more than nodes, memory as reconstructive rather than archival, augmentation understood through the higher rungs of Bloom’s taxonomy, and the questions of originality and transclusion that follow once documents become knowledge sculptures in spatial AR. The 21 July session deepened the philosophical layer, taking up navigational systems as a genuinely new category of tool, distributed cognition, the peer-to-peer movement of documents between people, and a critique of treating AI as a Cartesian oracle to be consulted rather than a system to be connected with — with Vannevar Bush, Doug Engelbart and Ted Nelson as the intellectual anchors they have long been for this community.
By 27 July the inquiry had become concrete and spatial: an experimental three-sphere “weave” view placing a chosen item at the centre with documents, people and locations arranged on concentric spheres around it — built explicitly to learn the interaction rather than to assert that the arrangement is right. The session also produced a deliberate terminological refusal: “citations” and “references” name a use, not the thing itself, and a knowledge space should treat artworks and other non-textual objects as first-class citizens. A complementary pipeline was shown running locally, with Apple Intelligence extracting visual media, abstracts and keywords from a PDF library, the chunking scaffolding built with Claude.
The community
July also enlarged the Lab. Patrick Lichty joined the sessions, bringing a background spanning digital art, VR, theatre and academic research, a Quest-and-WebXR practice that usefully complements the group’s Vision Pro work, and a long-standing interest in Roy Stringer’s Navihedron and in “thought molecules” — clusters of linked platonic-solid navigation structures. The regular voices — Mark Anderson, Peter Wasilko, Tom Haymes, Eric Rangell, Mohit Yadav, Peter Dimitrios, Brandel Zachernuk, with Dene Grigar present in the record — continued the Lab’s established rhythm of demonstration, challenge and synthesis, with the AI participating in the sessions as a named contributor whose outputs are labelled so they can be quoted without ambiguity. Each session, as ever, produced its analytical notes and its song.
Where this leaves us
The month closed with the design questions sharpened rather than settled: what belongs in a personal knowledge space, what its objects should be called, how they should arrange themselves around a person, and how documents should travel between people without platforms in the way. What changed is that these questions are no longer answered by argument alone. They are answered by building the thing by Thursday and finding out.