Media
2 items
Description
Highlights the tells of AI-generated prose on every page you read. Runs locally — no network requests, no data collected. Slop Finder highlights the tells of AI-generated prose on every page you read, and counts them on a toolbar badge. 38 patterns across three kinds of tells: - Vocabulary: "delve", "tapestry", "pivotal", "seamless", "ever-evolving" - Stock constructions: "not just X, but Y", "stands as a testament", "plays a crucial role", "no fluff, no filler, no jargon" - Structural habits: stacked rhetorical questions, echoing sentence runs, three sentences opening on the same word HOW IT WORKS - Exact matches get solid highlighter yellow; the containing sentence gets a pale wash so you can read the hit in context - The toolbar badge counts matches on the current tab - Built on the CSS Custom Highlight API: the page's DOM is never modified, so nothing shifts, breaks, or interferes with selection and copying - Re-scans as pages change, including busy feeds that render content late YOUR RULES The pattern set is plain JSON. Download the active set, edit it, upload it back — validation runs before anything is stored, and one click restores the defaults. Toggle the whole thing off from the popup at any time. PRIVATE BY CONSTRUCTION Everything runs locally in your browser. The only permission is "storage", used for your own settings. No network requests, no analytics, no accounts, no data collection of any kind. Detection patterns adapted from Simon Willison's LLM cliché highlighter. Open source: https://github.com/williamnewton/slopfinder