A browser buzzer that calls out corporate jargon while you talk. Say it straight or get buzzed.
Press the red button, talk like you're in a meeting, and every time you say "circle back" or "leverage synergies" you get a loud red BZZZT, a hazard flash, and the offending phrase slammed across the screen. Stop, and you get a Crap Score out of 100, a letter grade from A+ (Certified Human) down to F− (Synergy Final Boss), and a verdict you probably deserve.
TODO: demo GIF. Record a ~6 second capture of the buzzer firing on "let's circle back and take this offline" and drop it in as
docs/demo.gif, then embed it here. This is the single highest-value thing in this README — put it above everything else.
Live demo: TODO: add URL
TODO: screenshots. Add a shot of a Crap Score result card (the share image, 1200×630) right here. It's the thing people repost.
Your browser's Web Speech API transcribes your speech in real time. That transcript is matched
against a list of 3,600+ corporate jargon terms compiled into a single boundary-aware,
case-insensitive regex — every new hit triggers a buzz, a flash and a stamp. When you stop, the
buzzword density is turned into a 0–100 Crap Score, a grade, and a shareable result card rendered
client-side with <canvas>.
No build step, no framework, no dependencies. Open index.html — or, to get the microphone
working, serve it over localhost:
npx serveThe mic needs https:// or http://localhost. From file:// the typed fallback works but the
mic may be blocked.
Voice needs Chrome or Edge (Web Speech API). Firefox and Safari get a notice and the typed "rehearsal" input, which runs the same detector.
Your speech is transcribed by your browser and streamed to its provider (Google, in Chrome) to become text. Cut The Crap never stores or sends your speech. Analytics are cookieless and anonymous; the only personal data collected is an email address, and only if you type one into the waitlist form. Full detail in the privacy policy.
If this made you laugh, star the repo. And go get scored at TODO: add URL.
The list is the fun part and pull requests are very welcome. Edit jargon-source.md (one
- term per line), then run:
node scripts/build-jargon.mjsThat regenerates jargon.js, which is what the app actually loads. Editing jargon.js directly
works too. See CONTRIBUTING.md for what makes a good term and how not to break
the matcher.
| File | Role |
|---|---|
index.html |
the app — UI, detection engine, session logic |
jargon.js |
runtime term list (window.JARGON), generated |
jargon-source.md |
the human-editable term list |
score.js |
Crap Score, grade ladder, percentile — a pure function |
share.js |
result card canvas, share links, copy/intents |
analytics.js |
cookieless PostHog wrapper |
style.css |
the design system, shared by every page |
r.html |
the shared-result view (/r?s=…) |
privacy.html terms.html legal.html |
the legal layer |
api/waitlist.js |
serverless mailing-list signup endpoint |
This repo is the browser app, MIT licensed: the detection engine, the jargon dictionary, the scoring, the buzzer, the share card and the static pages. It runs entirely in the browser with no accounts, no database and no session storage, and it should stay that way.
PRs are very welcome for the jargon list, detection, accessibility and the UI. Please don't send
PRs that add a backend, authentication or data persistence — see
CONTRIBUTING.md. No credentials or .env files belong in this repo.
Static site on Vercel: import the repo, Framework Preset = Other, no build command, default
output directory. The one server-side piece is api/waitlist.js, which needs a
BUTTONDOWN_API_KEY environment variable. Every push to main redeploys.
MIT — see LICENSE. The name, wordmark, visual design and written copy are not covered by the MIT grant.