Andrej Karpathy vibe codes HackerTube video platform in two hours Other

AI researcher Andrej Karpathy launched HackerTube, a customized video recommendation platform curated from two decades of community discussions on Hacker News.

Karpathy built and deployed the full application in two hours over a weekend using Claude Opus 5.5 on Medium effort, automating data processing, interface design, and deployment configuration.

What it is

• Hacker News comment filtering. Karpathy tasked Claude with analyzing 43 million Hacker News comments spanning 20 years, distilling them into 6,454 high-signal educational and technical video recommendations.
• Two-hour build time. Using Claude Opus 5.5, the complete software pipeline was generated through prompt instructions. Karpathy reported that his manual actions were limited to purchasing the domain, configuring a YouTube API key, and provisioning a VPS server.
• Curated video directory. The resulting platform delivers a lightweight, ad-free interface focused exclusively on technical talks, computer science lectures, and engineering presentations favored by HN participants.
• Open dataset export. Karpathy released the processed comment analysis as a downloadable archive, allowing developers to prompt local AI agents to build custom recommendation feeds tailored to specific research domains.
• Expanding vibe coding scope. Karpathy emphasized that projects historically demanding weeks or months of engineering can now be executed autonomously within hours.

Project overview

Sources

• HackerTube Live Platform
• Andrej Karpathy on X