Software notebook / 2026
The Tesseract Memory.
A software notebook that keeps the decisions, failures, and changes close to the work.
- ROLE
- Design and implementation with Codex
- STATE OF THE BUILD
- In progress
- RECORD UPDATED
THE CONTEXT
What is being built.
The Tesseract Memory connects technical notes with the software they describe. The current site combines server-rendered reading pages, a compact build index, and a source-derived particle figure that follows the homepage’s real content sections. This website is also documented here.

The transparent Reach artwork master used by the static fallback and native Reach particle data. This is the source image, not a capture of the animated renderer.
CHOICES ALONG THE WAY
Why it took this shape.
- Make notes the primary record and use builds to provide context, rather than treating every website feature as a separate project.
- Keep navigation, headings, article bodies, and links outside the graphics layer.
- Separate source identity, generated-data correctness, and rendered appearance into distinct verification questions.
- Let the reader’s current scroll position determine the figure’s chapter instead of queuing or replaying transitions.
STRUCTURE & TOOLS
How it fits together.
- Server-side repository interfaces supply structured notes and build records to ordinary HTML reading routes.
- The homepage shares one particle renderer across responsive layouts; article pages do not mount it.
- Native scroll is mapped through measured content anchors, while the source artwork remains available during reduced motion or graphics failure.
MADE WITH
Next.js · React · TypeScript · Three.js
DATED & SCOPED
Evidence from the work.
Each entry describes what was checked or observed, and where. A local check is not a production claim.
- passed
Source identity checks
The reference and Reach files matched their recorded hashes and dimensions. Mutation, invalid-role, repository-path, and escaping-symlink cases were rejected. This verifies source identity, not rendered appearance.
ScopeLocal Node.js source-lock tests
- passed
Particle representation checks
Payload hashes, original Reach texel identities and RGBA, anatomical populations, four distinct pose frames, and encoded positions passed their deterministic checks. These are data checks, not a GPU performance or visual-fidelity claim.
ScopeLocal decoded-particle data tests
- passed
Native scroll and integration checks
Measured unequal anchors, small scroll movements, reversal, large jumps, invalid geometry, and uniform frame interpolation passed. Source checks confirm one shared figure and a passive scroll listener without page-scroll interception.
ScopeLocal scroll-math and source-integration tests
- passed
Continuous motion-field checks
Focused checks passed for protected anatomy, packet containment, local continuity and strain, and the original head–neck neighbour pairs across sampled poses and widths. These mathematical measurements do not establish GPU appearance or visual acceptance.
ScopeLocal CPU reference and motion-field tests
The source artwork shown above is part of this record. Its caption identifies the source and scope.
A NOTE TO TAKE FORWARD
What is still open.
The useful boundary is between the notebook, its evidence, and its visual enhancement. A source checksum cannot approve an animation, a smooth scroll controller cannot guarantee coherent anatomy, and an attractive page cannot substitute for a truthful record of the work.