Own HackerTube, a new AI-driven curation platform inspired by Andrej Karpathy’s HackerTube concept, has officially launched at ownhackertube.cc. The tool allows developers and founders to create personal, YouTube-style sites where an AI agent filters incoming videos based on user-defined standards rather than summarizing them. The project is in its earliest stages, offering a distinct alternative to the current wave of AI video summarizers by focusing on quality input over shorter summaries.

Filtering Over Summarization

The core philosophy of Own HackerTube is that better output requires better input. Unlike tools that condense video content into text, this platform acts as a gatekeeper. Users write their standards in plain Englishβ€”for example, rejecting clickbait or demanding concrete numbers in AI startup interviews. The AI then scores new videos from followed channels on a scale of 0 to 10, publishing only those that score 7 or higher. Each accepted video includes a brief explanation of why it was chosen and timestamps for key moments, which are sourced directly from the creator’s existing chapters, ensuring no invented metadata.

Infrastructure and Pricing

The platform runs on a daily batch processing model, checking new uploads every morning without requiring users to schedule tasks. The infrastructure supports YouTube and Bilibili (in beta), with the latter requiring a browser-sync workaround due to API limitations. For developers, the site generates an RSS feed and is indexed by search engines, making it a viable public portfolio. The free tier allows one channel and 30 AI checks per month, while the Pro plan costs $8 monthly or $59 annually, expanding the limit to 20 channels and 600 checks per month. Custom domains are currently rolling out for Pro users.

Key Takeaways

  • The AI scores videos from 0 to 10 and only publishes those scoring 7 or above.
  • Key moments are derived strictly from creator-defined chapters, preventing hallucinated timestamps.
  • Bilibili support is currently in beta and requires a browser-based sync method.
  • User feedback on picks directly trains the AI’s future scoring criteria.
  • The platform emphasizes 'watching' over 'reading summaries,' preserving the original media format.

The Bottom Line

Own HackerTube correctly identifies that the bottleneck for technical learners isn't reading speed but input quality. By filtering noise before it reaches the user, it restores the value of deep work in an era of AI-generated filler.