The v1 reuse library shipped and produced no measurable reuse gain
Closed in July. The library was used twice in six weeks, both times by its own author.
The record
The hypothesis was that a shared library would cut delivery time on repeat work. It did not, and the reason found at close was discovery rather than quality: nobody outside the authoring team knew what was in it. A v2 should start with how a component is found, not with what goes in it. Do not cite the reuse claim externally — it is not instrumented and the one measurement taken was negative.
Applies to
- Company
- AI Automation Agency
- Department
- Delivery
- Agent
- No single agent
Context bound to one project, from kickoff to close. Readable by: Everyone on the project.
Provenance
Reuse library close-out
Job run · unattended
logs/agency/job_1698.json
- Recorded by
- Delivery Manager
- First learned
- 22 Jul 2026
- Last updated
- 22 Jul 2026
Index state
Not embedded- Index
- project-archive
- Vector id
- —
- Model
- —
- Indexed at
- —
- Rank weight
- 1.00
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.