Slop Filter

At a glance

Slop Filter adds adjustable writing-style scores to posts and comments on X, LinkedIn, Reddit and YouTube, then badges, dims or hides items above your threshold. It is worth considering as a personal feed filter if repetitive prose bothers you enough to accept mistakes and send post text to an external model. Its score does not establish who wrote a post, so start by seeing whether its flags match your taste.

Good fit

Readers willing to tune what they see. TypeSafe AI’s Jev answers narrow style questions; local code combines those answers with punctuation and vocabulary cues, weighted rules and overriding signals. You can reveal hidden posts, label examples and fit weights to those labels. The appeal is control over your reading experience. Labels record your judgment and do not establish verified human or AI authorship.

Consider something else when

You need evidence for an authorship accusation, an offline filter, or a rule that never sends feed text outside the browser. Scores are unsuitable as proof for moderation or employment decisions. The account-blocking feature keeps a local extension list; it does not block or unfollow someone through the social platform itself.

Start here

The documented Chrome route is Developer mode and Load unpacked; Firefox uses a separately prepared temporary add-on. After reviewing the extension and choosing a provider, open its options and set “When a post is flagged” to “Only show a badge”. Enable “Label buttons”, inspect examples and compare the flags with your own choices before hiding posts or fitting weights. A useful result is a filter you agree with often enough to keep; detection accuracy has not been tested for this guide.

Cost and requirements

The code is MIT licensed, with separate font and icon licences. Scoring needs your TypeSafe or OpenRouter key, billed by that provider, or the developer’s “Use mine” service. Its September 26 terms list $3.99 per month, a card-required trial and automatic charging when the 100th post is scored. A previously used trial card or email can be charged at setup. Managed limits are 3,000 posts daily and 30,000 monthly. Read those terms before pairing an account.

Limits to know

The provider receives post text and relevant reply context. Keys and labels stay in local extension storage; exporting labels includes text and handles. The managed service’s policy says it does not retain posts, but keeps billing and usage records, with anti-abuse card/email hashes surviving deletion. Optional YouTube video scoring uses up to 150 words from English captions in the opening minute, not a whole-video assessment. Missing captions and service or site changes can prevent scoring.

Related work

• TypeSafe AI (context)

Still unclear

• How often the complete scoring rules wrongly flag independently verified human posts in your language and feed; the weight-fitting script evaluates your own labels, not verified authorship.