Omega — Universal Glossary Platform
One shared glossary per product — wired to the code — that tells you what a change will break.
Live in production; blast-radius review agent in pilot
- 3→1 — Three surfaces — a page, a REST API and an AI agent — behind one query engine that can't disagree
- 1,154 — Coupling links mapped across code and design for the reference product
- 0 — Cross-tenant vocabulary leaks — isolation is enforced in the database, not by convention
“If I change this, what breaks?”
Most teams cannot answer that without reading the code, and by then the change is usually written.
No one person holds the full picture of how the front end and the back end are coupled. Risk gets discovered after a merge rather than before it. Every product spells the same idea a different way, so AI tooling cannot reason across the mess. And the dependency diagram someone maintains by hand goes stale the moment a person forgets to update it.
What it answers
- Blast radius, in one query. Name a database collection, a service, a route, a screen or a provider, and get back every user flow that touches it — then everything those flows touch.
- A warning when something is load-bearing. Some artifacts fan out across most of the product. Rather than dump a hundred dependents on you, it marks them as shared and says to review carefully.
- Each product’s words kept apart. Isolation is enforced in the database rather than by convention: a non-member gets nothing, not a login-less peek.
- One source, three ways in. A page for people, an interface for build pipelines, and a server for AI agents — all reading through the same loader, so they cannot disagree with each other.
- An honest “not mapped”. Where nothing is linked yet, it says so explicitly instead of collapsing the gap into “no impact”.
Machine proposes, human disposes
A sync job keeps the map current: it adds new links, stamps each with how it was derived, and flags divergence.
What it will not do is overwrite a human’s work. It updates only the entries it wrote itself, never deletes, and leaves anything curated alone. The one step that stays manual is a developer declaring what a flow is — which is exactly the step where a human should be choosing the vocabulary.
Why “not mapped” matters more than it sounds
A tool that answers no impact when it simply has no data is worse than no tool, because the answer looks the same as a real all-clear.
Coverage gaps here are a first-class result, rendered differently and never collapsed to zero. You can trust a green because the amber exists.
Where it fits
- Teams whose front end and back end are coupled in ways nobody fully holds.
- Platform owners standardising vocabulary across several products.
- Reviewers who want a change’s blast radius before approving it.
- Teams building AI tooling that must reason across code, data and design in one shared vocabulary.
- Regulated work needing isolation and audit from the first commit.
Try it on something real
Stand up a space, seed your top five or six terms, and run the question against an artifact you already worry about.