DEMO / RECORDING
Overview & Result
Tony Dinh (@tdinh_me) developed an open-source Chrome extension that automatically identifies and skips sponsored promotional segments in YouTube videos in real time. Traditional LLMs take 2 to 4 seconds to parse transcripts, making real-time playback synchronization impossible. By leveraging Jev's sub-50ms bounded choice, the extension evaluates 5-second rolling audio segments on the fly, immediately seeking past sponsorships at an operating cost of roughly ~$0.005 per video.
How Jev fits in the loop
- Capture the active YouTube tab's audio stream or subtitle transcript via Chrome Extension background APIs.
- Chunk the spoken dialogue into rolling 5-second timestamped text buffers as the video plays.
- Invoke Jev Noul / Choice (<45ms) to determine if the active window contains sponsor outreach patterns.
- When a sponsor segment is confirmed, immediately execute a HTML5 video seek (video.currentTime = skipTarget) to resume normal playback.
How to reproduce
- Inspect author post and reproduction notes.
- Deploy agent loop and bind Jev decision endpoints.
- Run test inputs and observe latency and accuracy.
Why this build matters
Proves that lightweight, low-latency models can bring real-time semantic intelligence directly into consumer browser extensions without cloud lag or high API expenses.
Limitations
- Community observed build; metrics are author-reported.
- Requires third-party dependencies as described in source.