entry_id: RD-2026-005
project_id: PRJ-2026-003
core_activity: CA-01
title: Can software architecture become the primary view of a codebase?
timezone: Australia/Perth
period:
  started: 2026-07-11
  ended: 2026-07-14
research_effort:
  hours: 28
  basis: estimated
  note: Reconcile this public estimate with private contemporaneous company time records.
source_record:
  codex_session_id: 019f4fcf-b843-7222-b270-c23e8ec3a77a
  repositories:
    - mobily-enterprises/vibe64
    - mobily-enterprises/vibe64-online
  retained_artifacts:
    - graph-system.md design record
    - initial System world implementation in the isolated system-world-v1 worktree
  note: >-
    The complete conversation, intermediate source and screenshots are retained
    privately. This public record omits machine paths, customer material,
    credentials and private remuneration information.
live_evaluation:
  preview_handed_to_evaluator: "2026-07-12 08:27:03 +0800"
  representation_rejected: "2026-07-12 08:35:27 +0800"
  elapsed: approximately 8 minutes
  evaluator: product inventor
  decision: "Reject the architecture-first opening as too abstract."
hypothesis:
  recorded_at: "2026-07-11 14:21 +0800"
  statement: >-
    If framework-specific evidence can be compiled into a small common system
    model and projected through a constrained three-dimensional visual grammar,
    a systems-minded user should be able to identify responsibilities,
    interfaces, providers, consumers and structural risks without reading the
    implementation source.
primary_success_condition: >-
  Use Vibe64 itself to show its major subsystems, select a real operation, trace
  it to evidence-backed providers and consumers, expose uncertainty rather than
  filling gaps, and make at least one useful architectural question answerable
  without opening source.
controls:
  - Vibe64 was the single real subject rather than a purpose-built example.
  - The first adapter was limited to JSKIT while the common model remained framework-neutral.
  - Mechanically derived, declared, observed and inferred claims were kept conceptually distinct.
  - The System tool was scoped to the current working tree of an active session.
  - The user evaluated the live prototype before its visual assumptions were accepted.
result: rejected
reason: >-
  The extraction and model were technically viable, but the first client/server
  system world was judged too abstract during live use. It did not provide a
  sufficiently strong physical anchor for understanding where the architecture
  came from.
next_test: >-
  Keep the filesystem as the stable ground truth, encode files and directories
  directly, and add subsystem meaning as an evidence-backed overlay rather than
  making the architecture abstraction the initial world.
