- Pull the team's threads, docs, and meeting transcripts from the last 6 months.
- Read each to identify recurring acronyms, project nicknames, and internal jargon.
- Manually build a vocabulary list, deduplicating variants.
- For each term, read context to extract a canonical definition.
- Find two illustrative example interactions per term — skim and redact.
- Estimate usage frequency by counting occurrences.
- Flag any term with multiple competing definitions — surface conflicts.
- Compose the alphabetical glossary.
~90k tokens