Can an agent — or I — find where anything is defined?

Why it matters

An agent doesn't hold your whole repo in its head; it searches. If a concept can be written five different ways, search misses cases — and a confident agent that can't find the existing definition invents a second one. One shape per concept means search is complete, so the agent edits what's there instead of duplicating it.

See it

Glyph restricts mutation to two forms (mut x = … and mut x.method(…)), so every place state changes answers to a single search:

src/cart.glyph:14:   mut items.push(item)
src/cart.glyph:22:   mut total = total + price
src/user.glyph:8:    mut session.token = fresh

Three mutations, all of them, in one search. In TypeScript a mutation can hide behind =, .push(), Object.assign, a setter, or a spread — there's no single token to grep, so “does anything change this?” has no reliable answer. The same one-shape rule makes whole categories auditable:

# every UI component
grep -rn "^component " src/

# every function
grep -rn "^fn " src/

# the entire resource-ownership surface
grep -rn "owned" src/

There's no object-literal shorthand, no second way to declare a component, no overloaded syntax to disambiguate. Fewer ways to write a thing is a real constraint — and it's the point: it's what makes an agent's search, and yours, trustworthy.