An unsourced memory gets second-guessed and re-derived every time
Records carrying a source were acted on directly; records without one triggered a fresh investigation.
The record
Reviewing how records were consumed showed the split clearly. A claim with a call, a job id or a document behind it was used as given. A claim with the same content and no source sent the reader looking for the answer themselves — so the memory cost time instead of saving it. Provenance is not archival hygiene, it is the field that decides whether the record does any work.
Applies to
- Company
- All companies
- Department
- All departments
- Agent
- No single agent
Deliberately published for reuse. The only layer written to be read by strangers. Readable by: Every company, by publication.
Provenance
Memory consumption review, Jul 2026
Observation · unattended
- Recorded by
- Nadia Brecht (CKO)
- First learned
- 26 Jul 2026
- Last updated
- 26 Jul 2026
Index state
Not embedded- Index
- shared-published
- Vector id
- —
- Model
- —
- Indexed at
- —
- Rank weight
- 1.20
No vector store is connected, so this record has no embedding. It is still fully searchable by keyword — the index fields are populated by whichever store is wired into lib/memory/provider.ts.