Bouncer

Bouncer is a browser‑ and mobile‑based utility that intercepts the Twitter timeline and removes posts matching user‑defined filters. Users can enter phrases or abstract concepts such as emotions, and the system evaluates each tweet’s context, meaning, and tone before deciding whether to hide it. The tool also records which items are “bounced,” allowing the underlying recommendation algorithm to adjust over time and present a feed that aligns more closely with the user’s preferences.

The service is built on the open‑source language model Qwen 3.5 4B, which balances contextual understanding with the speed needed for real‑time scrolling. Currently the processing runs on the provider’s data center, but the developers are working on a version that can execute on the GPU resources of a laptop or phone, giving users control over where the inference occurs.

Bouncer is distributed as a Chrome extension, a Firefox add‑on, and an iPhone App Store app, and it is positioned as an experimental project aimed at giving individuals more agency over the content they see on Twitter.