index a06e2bb..5988b40 100644
@@ -16,4 +16,4 @@ an engine that takes a description and makes it more precise — cutting ambigui
the technical core is essentially an adversarial evaluation: given a description, can an LLM distinguish it from descriptions of adjacent things? if "a powerful note-taking app" describes 50 different products equally well, it's not a good description of yours. the engine would generate "imposters" — things that the description could also apply to — and rate precision by how well the description excludes them. this is a clean formalization of the problem: a description is precise if and only if it rules out alternatives. improving the description means shrinking the set of things it could apply to while still accurately describing the target.
-connects closely to [[writing-tools|writing tools suite]] which covers the broader space of text quality tools — ultimate-describer is the most conceptually novel piece of that suite. [[embedding-tone-interpolation|embedding tone interpolation]] is adjacent on the "understand what language is doing in embedding space" thread. [[personalized-autocomplete|personalized autocomplete]] shares the "AI that improves your writing in-context" framing. [[quality-search|quality content search]] is the search-side complement — if you can score description quality, you can filter for high-quality descriptions in search. [[idea-extraction-system|core idea extraction]] is the generalized abstraction version of the same capability: finding the essential structure of a thing.
\ No newline at end of file
+related: [[writing-tools|writing tools suite]], [[embedding-tone-interpolation|embedding tone interpolation]], [[personalized-autocomplete|personalized autocomplete]], [[quality-search|quality content search]], [[idea-extraction-system|core idea extraction]]
\ No newline at end of file