Add an organization-level /sanity/ architecture alignment view

[szmyty]

Outcome

Add an organization-level /sanity/ module to Organization Intelligence that lets the human architect periodically step outside the implementation loop, reconstruct the Ego Hygiene architecture as a whole, review an AI-assisted coherence assessment, and explicitly ratify or challenge the represented architecture.

This is the fleet-level counterpart to the repository /sanity/ view in egohygiene/relay#76 and should plug into the modular control-plane architecture from #27.

Primary user question

Do I still understand why this organization is shaped this way, do the repositories and boundaries make sense together, what is the AI worried about, and have I personally reviewed and accepted this current architecture?

The answer must not be derived from activity volume, issue count, CI status, or a model-generated score.

Architecture

Hygiene organization architecture + ownership policy
        +
ADRs + repository purposes + dependency evidence
        +
Aether coherence review (#87)
        +
Observatory normalized posture (#21)
        ↓
Organization Intelligence /sanity/
        ↓
human review + explicit ratification state

The organization UI remains a projection. Canonical architecture, review, and ratification evidence stay with their owning sources.

Default experience

Optimize for a human architect returning after days/weeks of rapid parallel work.

The initial screen should answer, without requiring 29 repository tabs:

• What are the major systems/repos and what does each own?
• How do they fit together?
• Which boundaries are especially important?
• What changed materially since I last reviewed the architecture?
• Are there new or contradictory dependencies?
• Are any responsibilities duplicated or unowned?
• Has complexity grown faster than demonstrated need?
• Which parts does the AI believe deserve human skepticism?
• Which architecture questions require my judgment?
• What exact architecture snapshot did I last ratify?
• Is that ratification still current?

Organization architecture map

Provide a zoomable/filterable architecture map or equivalent accessible presentation showing:

• repositories/systems;
• canonical responsibilities;
• dependency direction;
• consumes/provides relationships;
• organization contracts and major platform layers;
• important data/trust/execution boundaries;
• lifecycle/maturity where relevant;
• authoritative versus inferred relationships;
• links to repository /sanity/, /decisions/, /dependencies/, and canonical architecture sources.

The map must have a complete text/table equivalent.

Plain-language architecture explanation

Before deep diagrams, provide a concise explanation suitable for reorientation:

• the overall thesis;
• the major layers;
• what each major repository family does;
• how work/data/evidence flows;
• why the architecture is split this way;
• which complexity is intentional;
• which areas remain provisional.

This should be optimized for "I have been speedrunning for a week; remind me what the hell all of this is."

Reviewer/challenger assessment

Render the latest Aether architecture-coherence review separately from human disposition.

Include:

• strengths;
• concerns;
• duplicated/unclear ownership;
• questionable dependency directions;
• missing canonical owners;
• complexity/fragmentation concerns;
• architecture-versus-implementation drift;
• unresolved assumptions;
• credible alternative architectures/counterarguments;
• recommendation and confidence;
• evidence freshness.

The reviewer should be encouraged to challenge the existing architecture rather than merely explain why it is internally consistent.

Human ratification

Display an explicit organization-level human state such as:

• not reviewed;
• review due;
• review in progress;
• ratified;
• ratified with notes;
• revision required;
• rejected;
• deferred;
• stale.

A ratified state must identify the exact represented organization/repository architecture revisions and review record.

The organization page cannot infer ratification from merges, commits, closed issues, or continued development.

Changes since last review

Prioritize material architecture changes only, such as:

• newly created/retired repositories;
• changed canonical ownership;
• accepted organization/repository ADRs that affect topology;
• new cross-repository dependencies;
• extracted/shared capabilities;
• new persistent data/trust/execution boundaries;
• major changes to the platform thesis/distribution model;
• previously provisional pieces becoming canonical or deprecated.

Routine implementation noise should stay out of the primary view.

Review queue

Provide a short human-decision queue derived from unresolved architecture questions, for example:

• Should these two repositories remain separate?
• Is this capability owned in the right place?
• Does this dependency violate the intended direction?
• Is this abstraction justified yet?
• Did rapid iteration create a duplicated platform?
• Is a repository now obsolete or missing?

Each question must link to evidence and remain proposed until the human records a disposition.

Relationship to Organization Intelligence

Cross-link rather than duplicate existing modules:

• /intelligence/ overview / Now;
• Hygiene posture;
• Audits posture;
• Decisions;
• Dependencies;
• Roadmap;
• Releases/distribution;
• per-repository /sanity/ views.

The organization /sanity/ view is the human architecture alignment checkpoint across the fleet.

No universal score

Do not render a single "organization sanity = 92%" score.

If compact indicators are useful, keep independent evidence-backed dimensions such as:

• ownership clarity;
• dependency coherence;
• architecture coverage;
• complexity justification;
• unresolved decisions;
• review freshness;
• human ratification.

Unknown and not-applicable states must remain possible.

Privacy

• Public organization views must not expose private repository topology or evidence.
• Private/local builds may include authorized private organization snapshots.
• Visibility filtering occurs before rendering/aggregation.
• Human review records must not expose sensitive personal information beyond repository governance conventions.

Acceptance criteria

• Organization Intelligence exposes a /sanity/ architecture alignment module.
• The module consumes normalized Observatory evidence and Aether/Hygiene contracts rather than inventing another model.
• A plain-language organization architecture explanation is available.
• Repository/system ownership and dependency direction are navigable at organization scale.
• AI/reviewer assessment and human ratification remain visibly separate.
• Material changes since last ratification are prioritized over implementation noise.
• A short unresolved human architecture decision queue exists.
• Organization and repository /sanity/ views deep-link to one another.
• No opaque numeric architecture score is used.
• Never-reviewed, stale, revision-required, deferred, unavailable, and ratified states render explicitly.
• Every summary links to represented revisions and canonical evidence where authorized.
• Public/private visibility boundaries are enforced before aggregation.
• Mobile, keyboard, screen-reader, reduced-motion, and text-equivalent architecture views are verified.

Dependencies / related

• Modular Organization Intelligence: Evolve Organization Intelligence into a modular fleet control plane #27
• Existing portal: Build an organization-level Intelligence portal at egohygiene.io/intelligence/ #15
• Aether coherence workflow: egohygiene/aether#87
• Hygiene ratification policy: egohygiene/hygiene#59
• Observatory posture: egohygiene/observatory#21
• Repository /sanity/: egohygiene/relay#76
• Portfolio/focus map: egohygiene/pace#25

Non-goals

• Letting the AI approve the organization architecture.
• Making the dashboard the canonical architecture source.
• Automatically merging, deleting, splitting, or creating repositories from review findings.
• Treating every complex architecture as over-engineered.
• Hiding uncertainty behind a score.