precision description engine

an engine that takes a description and makes it more precise — cutting ambiguity, sharpening specificity, identifying where vague words are doing work that concrete ones should be doing. the motivation: most writing (product descriptions, essays, emails, technical specs) is precise enough to feel real but vague enough to mean nothing. "a comprehensive solution for modern teams" describes everything and nothing. the tool pushes back on this by flagging weasel words, suggesting more specific replacements, and scoring descriptions on how distinguishable they are from alternatives. this project won a hackathon.

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.