Canonical science
Registered TeX
Physics and derivational claims live in registered TeX sources and their claim-boundary context, not in search mirrors or generated notes.
AI research-agent deep dive / Phase 5C
Memory And Registries
AEther Flow uses memory, registries, wiki notes, semantic extracts, and local retrieval surfaces to find source evidence quickly. Their purpose is navigation and orientation. They do not replace tracked source files, registry rows, or source authority.
Query workflow
Memory preflight is a disciplined navigation step.
The pattern is two-stage: use memory status and targeted lookup to find likely evidence, then inspect the named canonical source or registry row before relying on it.
Status
Check freshness
Memory status and freshness warnings reveal retrieval drift. A warning is a maintenance signal, not a new source claim.
Lookup
Find likely paths
Targeted lookup or search helps locate object IDs, registry rows, prior tasks, or likely source paths without broadening the task.
Inspect
Open canonical sources
Any memory hit that affects routing, claim language, source selection, or project-control changes must be checked against source files or registry rows.
Receipt
Record evidence
New control transactions that use memory preflight record status, queries, returned IDs, inspected paths, registries, and hashes.
Freshness and drift
Stale retrieval is a maintenance signal, not authority.
Retrieval drift can matter operationally because it may hide current source evidence. It still does not change the source hierarchy or create a project-control claim.
Stale mirror
Refresh support layers
If a raw mirror, Obsidian note, semantic extract, or SQLite index lags tracked files, refresh retrieval layers and inspect the source directly.
Generated disagreement
Source wins
If a wiki note, generated HTML page, PDF, or public summary disagrees with registered source material, the derivative needs repair.
Validate-only
Check without writing
Validate-only commands can show generated-state problems, but they do not refresh stale derivatives or promote generated artifacts.
Local cache
Never cite as authority
.local prompts, previews, run caches, and local retrieval indexes are convenience state, not committed evidence or scientific authority.
Reader path
Read memory with workflow, roles, and source authority.
Memory and registries connect the AI workflow to source evidence: workflow says when memory is used, roles and skills say who may use it, and source authority says what must be inspected before claims move.
Back to AI research-agent trackReturn to the AI track landing page for the deep-dive page family.Research-agent workflowRead how memory preflight fits into bounded Director decisions and AgentJobs.Roles and skillsRead how the memory skill remains subordinate to source authority and task-local permissions.Source authority pageReview the broader trust map for website pages, generated derivatives, registries, and tracked sources.Upstream memory explainerGenerated noncanonical source-backed explainer for memory, registries, wiki, and retrieval surfaces.Upstream source authority explainerGenerated noncanonical source-backed trust map for generated derivatives and authority layers.
Source authority
What this page can and cannot establish.
This page can orient readers to source-first memory, registry rows, wiki notes, semantic extracts, Obsidian mirrors, SQLite indexes, and local caches. It cannot change memory behavior, change registry schema, override tracked files, promote generated derivatives, or establish physics or workflow authority.