Technical diary · generated from gaad_diary_technical_20260726.md by Scripts/gaad_diary_html.py · never hand-edited
The genome programme went from format to proven lineage in a single day. The bench ran five versions, v3_030 through v3_034, each confirmed by Miles before the next began; the spec ran v0.009 to v0.010 with a full flag-and-fold cycle; and the platform ended the day wearing the clean six-family face ruled the day before.
THE CONFORMANCE PASS - the D202 deliverable (P074)
A web-written genome (afal_hyper_g-c73a02, HYPER, 420 instances) was imported by the Blender host and rebuilt exactly. Check Conformance compared Blender's canonical float64 derived set against the web's emitted set, unrounded, within the D206 tolerances.
CONFORMANCE PASS, worst 7.1e-15 m, matching the bench pre-check digit for digit. x/y worst 8.9e-16 m (one float64 ULP at 8 m, the sin/cos library difference D206 was sized for); orientation and module scale exactly zero.
The tolerance sits five to six orders above the measured noise and six below the millimetre that matters. Sized, not guessed, and now hit.
DESCENT RULED AND SHIPPED (D209, P072, P073)
The identity gap: minutes after the PASS, an active-gene edit to the imported specimen left it wearing g-c73a02's id while no longer being that form. An Export then would have written a different form under the same id.
The ruling: a hand-edited specimen is a CHILD, not a new root. Parent, generation+1, lineage_id carried through. Unit of descent is the exported child, not the slider event. Only active-gene edits count (GENOME_ACTIVE_GENES is the one truth table for writing, greying, invalidation and descent). Species is not a boundary: a species change is still descent on the same lineage_id, marked crossed.
Spec v0.010 folded the ruling; the bench flagged F1-F6 and all were folded (F1 Export stores the genome on the object, giving descent one stash source; F3 species-flip host divergence accepted as UI behaviour; F4 descent lives at lineage.descent; the rest confirmed or corrected).
v3_033 built the mechanics on both hosts in the same release.
THE FIRST LINEAGE - L-4836 (P072, P073)
L-4836
g-c73a02 HYPER root gen 0 glas-web
g-237c0b PHYLLO crossed gen 1 glas-blender
g-2e2223 SHELL crossed gen 2 glas-blender
g-edb7c6 SHELL hand gen 3 glas-blender
Every v0.010 mechanism exercised except bred, which correctly waits for the breeding spec. Heritage genes (radius 8, rise 0.1) read straight from the files down the chain.
THE CLEAN AFAL STUDIO PANEL - v3_034 (F054, D208)
D208's platform IA shipped and confirmed live: SPECIES / GENERATE / PRESETS / PARAMETERS / RNG / GENOME FILE. Killed per the rulings: the 144 world, the 1994 Growth Generator, DEMO, PRESENTATION, Test D197, the mode row, the Gravity toggle, the Module picker (fixed to Cube), BREEDING and MEMORY. Retired panels unregistered, code kept whole.
Confirmed by a full browse, generate, and genome export/import/refusal sweep through the clean face.
Conscious costs on the record: Spiral, Cone and Hyperboloid still roll from the internally pinned shared envelope until their per-species windows land.
DRIFT CAUGHT TWICE MORE (T032)
Web z-law: the preview carried an invented Hyperboloid vertical law (z = u*radius*2.4); engine truth is z = index*Rise for every species. Fixed - the web conforms, the canonical evaluator is the single law in that host.
Blender inert slider: an unused gene slider presented as live while doing nothing. Fixed with per-species greying, visibly confirmed on the re-imported child g-237c0b.
Also fixed this release: glas_version now derived from bl_info; the v3_032 over-invalidation, by consulting GENOME_ACTIVE_GENES.
RECORDS FILED TODAY
Decisions: D209 (descent).
Pivotal moments: P072 (first descent, updated to the L-4836 chain), P073 (descent ships, both hosts one release), P074 (the conformance PASS).
Features: F054 (clean AFAL Studio panel).
Tickets: T032 (web z-law + Blender inert-slider relays), T033 (the Language file path evaporates in unsaved sessions).
Format: genome format v0.010 with the descent field and the active-gene invalidation rule.
DATE NOTE
The 25/26 Jul boundary blurred across sessions (v3_030 built late on the 25th, confirmed the 26th). Per the bench, 26 Jul is the date of the conformance PASS, the first lineage, v0.010, and the v3_031-v3_034 confirmations. Genome provenance dates are machine-written and correct.
OPEN ACTIONS
AFAL
A0019 high / Claude - the language file is behind the engine (gates the exporter; now also T033)
A0024 high / Claude Code - per-species ranges: PART DONE (Shell + Phyllo done; Spiral, Cone, Hyperboloid remain on the shared envelope)
A0020 medium / Claude - T1 research for the bench species
A0021 medium / Claude - surface topology: the tier that makes a panel system possible
A0027 medium / Miles + Code Technician - itemise the v3_029 issues Miles saw in Blender
A0025 low / Claude Code - geometric NATURE row on the Singularity panel
Films
A0017 medium / Miles - Film 001 - The Rebirth: storyboard v3, real GAAD footage: PART DONE
GALF
A0011 high / Miles - rule the fourth element: does a relationship carry a rationale
A0012 high / Miles - TWAL s6: map the three retired types
A0013 medium / Miles - ARAL s6 declares types but states no triples
MHAL
A0001 high / Miles - Midtown East recount under the D181 tax block rule
A0015 medium / Miles - the prototype carries MHAL verbatim at v0.003
Platform
A0014 high / Claude - the language-file generator: JSON to HTML, in the pipeline: PART DONE
A0022 high / Claude - AFAL has no assert contract and no Validate steps
A0016 medium / Miles - Fly Past on generator strings: PART DONE
A0008 low / Miles - Prototype/ carries stray ARAL versions
Website
A0007 medium / Claude - canonicals for generated record and language pages: PART DONE
A0023 medium / Claude - Episode 6: AFAL and the singularity
A0006 low / Miles - latest episode first: name the surface
A0010 low / Miles - website drifts from the bundle - accepted, not overlooked: PART DONE
STILL OPEN AT CLOSE
Standing threads for the bench (no order ruled): the conformant engine rebuild (float64 Python drives the nodes, ENGINE_VERSION 9); per-species window rollout for Spiral, Cone and Hyperboloid; the Language path to AddonPreferences with a .json filter (T033); v3_029's phyllo issues (never itemised, A0027); the web A0019 catch-up (pin language.version from the file); breeding (parked, with bred, its rng, and the parent model).
Diaries are held and await Miles's red pen; they travel in the diaries zip (D159), never the live bundle.