Slop-Meter: An Anonymous AI-Slop Rating Widget
A small, reusable, privacy-first widget that lets readers anonymously rate how much any piece of writing smells of AI slop. A slider from human craft to pure slop, with a live crowd verdict.
Slop-meter is a small side project that started with a simple itch. I wanted to know when readers feel that something I’ve written reads like AI slop (formulaic, padded, technically fine and completely lifeless), and I wanted that signal to be anonymous, a little bit playful, and impossible to ignore.
The result is a slider you drag from human craft to pure slop. Let go, and it records your verdict and shows you how the crowd scored the same piece. There’s a live one at the bottom of every post on this site.
What it is
Two decoupled, reusable parts plus an owner dashboard:
- A Cloudflare Worker API (the brain). Serverless, backed by D1 (SQLite), keyed by an opaque id so one deployment can serve many sites.
- An embeddable web component (the face). A single dependency-free
<slop-meter>element that drops into any page with one script tag and adopts the host theme. - A token-gated dashboard. An owner-only view of which posts have been voted on, their average slop score, and which ones to go and look at.
Design principles
- Anonymous. No accounts, no cookies, no personal data stored. The only identifier is a salted, daily, per-post one-way hash used purely to stop casual double-voting.
- Engaging. A continuous slider with live labels and an emoji burst beats a dreary thumbs-up/down.
- Reusable. Built as a standalone function from day one, MIT licensed, ready to lift into its own repository so anyone can deploy it.
Where it’s heading
Each vote is now bound to a version of the content it rated (a hash of the post’s markdown, captured as a one-per-version snapshot), so as I revise a post I can see how the slop rating moves, and pair each rating with the exact text that earned it. That’s the beginning of a feedback loop: use what reads as slop to write better.
Building that surfaced a clean per-post markdown endpoint, which turned out to be the seed of a separate experiment: making this site agent-readable.
The next step I’m interested in is exposing the read-only ratings data to an AI assistant, via MCP, so I can ask for insights and get a live, generated dashboard back rather than reading a table.