Do not start with the entire company. Start with one bounded system.
Good first candidates:
- one application with a few repos
- one service family with supporting docs
- one product area with support content and code
For a meaningful first test, aim for:
- 1 to 3 repositories
- 5 to 20 docs or support articles
- one local Obsidian vault path
- one local workspace path
You do not need everything on day one. A successful first run means:
- the vault is scaffolded and navigable
- research notes preserve provenance
- there is a linkable path from documentation to code
- the readiness report tells the truth about what is done and what is missing
- you can explain what should become automated next
After the first run, try asking:
- What are the main product areas and how are they represented in the vault?
- Which repositories implement this area?
- What documentation appears stale versus the code?
- What are the strongest bug or support signals?
- What is still missing before engineers can ship confidently from this system?
- skipping the manifest
- starting with too many repos
- importing notes without preserving raw sources
- treating the vault like a dump instead of a synthesis layer
- trying to automate everything before the manual path is proven