Why component names matter more with AI in the loop
Models select components largely from their names and prop signatures, without opening the implementation. Names that state purpose get chosen correctly; generic or abbreviated names get chosen by guesswork. Naming has moved from a readability concern to an interface concern.
Naming has always been in the "hard problems" joke, treated as a matter of taste and readability. With a model choosing components from a manifest, it stopped being about taste.
A human faced with Box2 opens the file and finds out. A model reads the name, reads the props, and decides.
Assume the implementation is never read
This is the mental shift. The name and signature are not a summary of the component; for selection purposes they are the component.
Which means every ambiguity in a name is resolved by guessing, and guessing is resolved in favour of whatever is most common in general, not most correct in your codebase.
Purpose, not position or appearance
TopSection describes where it sits, which changes. BlueCard describes how it looks, which changes at the next brand review. PlanComparisonTable describes what it is for, and that survives both.
One concept, one name
Codebases accumulate synonyms: Modal, Dialog and Overlay, all present, subtly different, nobody sure which is canonical. Humans muddle through by asking someone. A model picks whichever the name suggests and moves on.
Consolidating synonyms is unglamorous work with a direct payoff in generation quality — and it makes the library easier for new engineers too.
Props are part of the name
A prop named variant typed as a string is an invitation to invent values. The same prop typed as a union of four literals is a constraint.
Booleans deserve the same care: compact says what it does; mode2 says nothing and will be passed at random.
Fix what actually breaks
None of this requires a renaming project. Watch which components get selected wrongly — the list is short and stable — and rename those.
The pattern is usually obvious once you look: the abbreviated ones, the numbered ones, and the two that mean almost the same thing.
Frequently asked questions
Is not a long name worse to type?
Autocomplete makes typing a non-issue, and the cost of a wrong component reaching review is much higher than a few extra characters. Optimise for the reader, human or otherwise.
What about existing badly named components?
Rename the ones that get picked wrongly, which is a much shorter list than renaming everything. Track which mistakes actually recur and fix those; leave the rest.
