Bouncer, a new browser extension, lets users block unwanted content from their X/Twitter feed using AI. Describe what you don't want to see in plain language — "crypto", "engagement bait", "rage politics" — and Bouncer classifies and hides matching posts in real time. Available for Chrome, Edge, and iOS, the extension uses a MutationObserver to watch your feed, extract post text and images, then sends them to your chosen AI model for classification.

The tool supports multiple AI backends. Users can run models locally via WebLLM — Qwen3-4B, Qwen3.5-4B, and Qwen3.5-4B Vision run entirely in-browser with WebGPU, sending zero data externally. Alternatively, connect cloud APIs from OpenAI (GPT-5 Nano), Google Gemini (2.5 Flash Lite, 2.5 Flash), Anthropic (Claude Haiku 4.5), or OpenRouter (Nemotron Nano 12B VL, Ministral 3B). Multimodal models can even filter posts based on images, not just text.

Bouncer shows reasoning for why each post was filtered and caches results to avoid redundant inference calls. The filtered posts list lets users review what was hidden and why. This gives X users a personal content moderation layer the platform itself doesn't offer — filtering based on individual preferences rather than relying on Twitter's algorithmic decisions.