Get help, report an issue, or ask about a feature.
Bug reports, feature requests, false-positive complaints, "the model isn't loading", anything else — send it to the address below and I'll get back to you. SlopRadar is built and maintained by one person, so emails reach me directly.
slopradar@featureboard.ai →SlopRadar uses Chrome's built-in Gemini Nano model. It needs to be enabled in your browser. Open the extension's settings, look at the Status panel, and click Restart Engine. If the status stays "unavailable", make sure you're on a recent version of Chrome (newer builds ship Gemini Nano on capable hardware).
You can check Chrome's own status by visiting chrome://on-device-internals. If the model is downloading, give it a few minutes — it's a one-time download and runs offline afterwards.
Two likely causes. First, the default confidence threshold is 80% — if you'd like fewer false positives, raise it in Settings → Sensitivity. Second, the model is making genuine judgments about each post, and on a feed with a lot of marketing posts it'll flag a lot of them. Click "Not slop" on any incorrect flag and SlopRadar adds it to your personal patterns; the filter learns over time.
Gemini Nano runs on your device, which means classification speed depends on your hardware. On many machines it's about 2-5 seconds per post; on slower machines it can be 8-10 seconds. SlopRadar pipelines up to 3 posts at once, but the model itself is serial. Posts in your current viewport get priority — so scrolling away from a post may push it off the queue.
Yes. Right-click any text on a post and choose "Mark selection as slop". SlopRadar adds that fingerprint to your personal pattern list — any post matching that exact phrase will be hidden instantly on every supported site, no AI call needed.
Those two are in beta because the DOM selectors haven't been hardened against every layout variant yet. Core detection works, but if you see a post that SlopRadar should have flagged but didn't, please email me with the URL and a screenshot.
Not yet. SlopRadar relies on Chrome's built-in LanguageModel API, which currently ships only in Chromium-based browsers with Gemini Nano support. Edge and Brave (recent versions) should work; Firefox and Safari don't expose an equivalent API yet.
This isn't built in yet — your patterns and verdict cache live in Chrome's extension storage local to your profile. If this is something you need, email me and I'll prioritize it.
SlopRadar is free. If you find it useful and want to support continued development, the best thing is to tell people about it, file good bug reports, and consider FeatureBoard if you ever need help applying small ML models to a problem at your company.
The most useful bug reports include:
Send it all to slopradar@featureboard.ai.
SlopRadar is built and maintained by Lewis Valentine, an independent developer and AI consultant. It's a sister product to FeatureBoard, but uses no shared infrastructure — see the privacy policy for the full scope.