Best for
- Teams discovering in month three that 'active user' meant four things
- Onboarding into a vocabulary that currently lives in veterans' heads
- The domain-heavy project where business and engineering need one language
What you give it
- The documents and conversations where the terms live (specs, tickets, meeting notes), and the arbiters for genuinely contested terms
What you get back
- The term inventory, triaged: the dangerous ones first — terms used with multiple meanings by people who do not know they disagree
- Definitions that hold: each tested against real usage, with what the term does NOT cover stated, the near-synonyms disambiguated, and the contested ones resolved by a named decision rather than a quiet averaging
- The living apparatus: the glossary where work happens, new-term intake, and the drift check that keeps definitions matching usage
How it works
- Mines real usage, not memory: the terms as actually used in specs, tickets, dashboards and calls — because the dangerous ambiguity is the one people do not know they have.
- Triages by danger: the terms with multiple live meanings and consequential decisions riding on them come first — a glossary of eighty harmless terms burying eight dangerous ones protects nobody.
- Defines to hold: each definition tested against the found real usages (does it classify the edge cases?), the NOT-clause stated, the near-synonyms explicitly disambiguated — and genuine contests resolved by a named arbiter, because committee-averaged definitions please everyone and pin nothing.
- Installs the living layer: the glossary where the team already works, the intake for new terms, and the periodic drift check — a glossary is a claim about usage, and usage moves.
Example
You: Build the glossary for our logistics platform; sales, ops and engineering each speak a different dialect.
Result: The sweep of specs, tickets and call notes surfaced 60 candidate terms; the triage found the eight dangerous ones — 'shipment' topped the list (sales counted orders, ops counted pallets, the API counted container-movements; three dashboards disagreed because of it — the term got three precise names and one decision about which owns the bare word). Each definition shipped with its NOT-clause ('a delivery attempt is not a delivery') and its near-synonym map (consignment vs shipment vs order — the diagram that ended the Tuesday argument). The contested two went to named arbiters instead of committee-averaging. Apparatus installed: the glossary lives in the wiki the team actually opens, new terms enter through the same triage, and the quarterly drift check found its first divergence (a new team using 'hub' for something else) before it propagated. The month-three meeting where everyone discovers they disagreed: cancelled, permanently.
Limits — please read
- Pinning a contested term is a decision someone must own; the glossary frames it and records it, authority makes it.
- A glossary ends vocabulary ambiguity, not strategy disagreement — sometimes the word fight was a proxy, and the build says so when it finds that.
- Adoption lives in the plumbing: definitions linked where the terms appear (the dashboard, the spec template) beat definitions filed where nobody looks.