KoalaTube: Yelp for YouTube videos that maximizes Return on Attention

Transcript-aware YouTube ratings that help you spend attention where it pays back.

KoalaTube helps you assess YouTube videos with local caption estimates and your own feedback. On the watch page, see a caption-based estimate with a criterion breakdown, then mark the video Worth it, Okay, or Skip. Video-card badges distinguish your direct rating from the caption estimate. Channel feedback is shown only for matching channel links; a display-name-only match is labeled unverified, with personal fit unknown. Adjust the rubric weights, relevant topics, and low-information keywords in settings. The popup shows recently observed videos and a local activity summary, with links to revisit videos and a Reset control. How it works KoalaTube uses rules running in your browser to analyze available captions and video/channel metadata. Estimates are heuristics, not factual verification or calibrated predictions of personal relevance. Captions may be unavailable, and YouTube layout changes can affect the interface. Data handling While active on YouTube, the extension observes video identifiers, titles, channel metadata, playback progress, and timestamps. It also processes displayed video cards, your explicit feedback, and your rubric preferences. Local records and derived caches use browser extension storage. Retrieving captions requires requests to YouTube and caption URLs supplied by its responses. Requests can include your existing YouTube browser session; the player request forwards a YouTube-provided visitor identifier. This version has no developer-operated data upload or analytics endpoint. It does not request or import your YouTube account watch-history list. Use Reset to clear KoalaTube's stored state. Continued YouTube use can create new local observations; disable the extension to stop its activity. What's new in 0.1.2 This update separates caption estimates from explicit feedback on video cards. Channel-link matches and unverified name matches are labeled distinctly. Badges use a static metadata position with complete wrapping labels and accessible descriptions. Navigation and stop cleanup remove old badges and cancel pending caption work. Local storage writes are coordinated to protect feedback during concurrent updates. Caption requests have bounded deadlines, and derived caches have maintenance limits.