Jev Lens is the extension I built after getting tired of scrolling past noise to find the three posts a day that matter. It badges X and LinkedIn posts READ, MAYBE or SKIP in about 300 milliseconds each, using Jev, a System 1 model that classifies instead of generating. Install it unpacked from the repo, add your key and your interest profile, and the feed starts sorting itself.
Feeds are mostly noise. Keyword filters cannot tell a build log from a certificate announcement, and asking a language model about every post is slow and expensive.
Jev does not generate text. It classifies, fast, against a profile of what I care about. The extension watches the feed, sends each post as it appears, and paints a badge on it in real time. Everything runs with your own key; nothing of ours sits in the middle.
A feed that triages itself. Episode 15 of The Agentic Engineer shows the model behind it and why System 1 models fit this job; the companion blog post covers Jev and RLCD for testers.