new file mode 100644
index 0000000..7ae08d8
@@ -0,0 +1,33 @@
+---
+type: idea
+title: precision description engine
+tags: [writing, ai, nlp, software]
+status: built
+first_captured: 2026-03-15
+updated: 2026-04-10
+sources:
+ - sources/ideaflow/2026-03-15_framing-generally-totally-matters.md
+ - sources/google-sheets-ideas.md
+---
+
+# precision description engine
+
+"ultimate describer" — an engine for making descriptions more precise. won a hackathon. the architecture is layered:
+
+1. **core engine** — precision mapping: take input text, identify vague/ambiguous terms, generate a spectrum of replacements from vague to precise, score along a specificity axis
+2. **context calibration** — audience profiling, domain vocabulary, cross-language precision, emotional register
+3. **personalization** — quiz/preference system, style inference from user writing, comparative selection
+4. **output modes** — inline text replacement, visual generation, vague-to-precise slider
+5. **applications** — anti-slop writing, copywriting, job/program applications, [[referral-hiring|hiring]] role descriptions, VC pitches, self-intros at events, teaching users to describe better, Claude skill writing, media generation, translational meaning preservation, writing in someone's voice
+6. **evaluation** — A/B testing, user inspection, iterative model improvement
+
+the core insight that "framing generally totally matters" — at networking events, how you describe yourself (timid vs boisterous vs precise) completely changes the outcome. this tool systematizes precision in self-description. connects to [[oncue|OnCue]] for the interview/application output side, and to the [[cultural-translator|cultural translator]] for cross-context communication.
+
+**spreadsheet evaluation:** excitement 10/10. the spreadsheet contains an extensive application list beyond the hackathon version: college supplemental essays/scholarships/awards, teacher feedback and learning optimization, young founder cold outreach, grant/proposal writing (noted competitor: GrantBoost), sales, event invitations, difficult conversations and negotiations. tech depth ranked 4/10, labeled [WRONG FIT] ("useful but thin technically — more of a prompt engineering project than a hard problem"). the shenal-maker/ultimate-describer GitHub repo and hackathon win are documented.
+
+---
+
+## timeline
+
+- [2026-03-15] captured with full layered architecture — won a hackathon
+- [2026-04-10] google sheets evaluation — 10/10 excitement, expanded application list, [WRONG FIT] for tech depth