Karpathy Built the HackerTube Video Recommendation Website with Claude Code in Two Hours

Disgusted by YouTube's poor recommendations, AI guru Karpathy "hand-built" a replacement site in 2 hours. Phoenix Tech Download the app for exclusive early access. November 02, 2026 13:45:38 From Beijing. Updated November 02, 2026 13:47:53

Phoenix Tech News, November 2 — Andrej Karpathy, former Tesla AI director and OpenAI co-founding member, recently released a weekend project called HackerTube. Displeased with YouTube's algorithmic recommendations, he used Anthropic's AI coding tool Claude Code to filter 6,454 repeatedly recommended videos from 43 million comments posted over the past 20 years on the tech community Hacker News, and built a brand-new video recommendation website. The entire process took only about 2 hours.

Karpathy said on social media that compared with YouTube's existing recommendation mechanism, HackerTube can provide a "significantly better viewing experience."

Unlike video platforms that rely on users' viewing history and personalized recommendation algorithms, HackerTube's core idea is to use real community discussion records to find video content that has been recommended and recognized over the long term.

The project's data source is comments from Hacker News over the past 20 years. Karpathy had Claude analyze the 43 million comments, ultimately filtering out 6,454 videos that were continuously recommended, thereby building a curated video library.

HackerTube also has several designs that differ from traditional video platforms. First, users do not need to register or log in to an account; usage status is stored directly in the local browser session. Second, the site no longer uses the cover images uploaded by YouTube video creators, but instead captures a representative frame from the middle of the video as the thumbnail.

Karpathy believes this can avoid video covers that have undergone repeated A/B testing and been specifically optimized to attract clicks, letting users focus more on the content of the video itself.

In addition, HackerTube also provides a ZIP archive of the underlying comment analysis data, allowing other developers to download it, hand it to Claude for analysis, and create different versions of video recommendation websites according to their own needs.

More noteworthy than HackerTube itself is the development process Karpathy disclosed.

According to him, the entire project was mainly completed using Claude Code, adopting the Claude Opus 5.5 model with reasoning effort set to Medium. He used the Mac version of the Claude Code desktop app.

Karpathy said he only needed to converse with Claude to let the AI complete almost all development work.

Throughout the process, he personally handled mainly only three operations: creating an API key through the Google Cloud interface to query YouTube video metadata; starting a VPS server for deploying the website; and purchasing the website domain.

After the server was ready, Karpathy provided the IP address to Claude and asked it to update the code, strengthen security protection, complete deployment, and optimize performance. According to him, Claude then completed these tasks.

He finally raised a question: "What other possibilities have we not yet thought of?"

HackerTube is not only a redesign of YouTube's recommendation mechanism, but also demonstrates that AI coding tools are changing the way individual developers build internet products. However, the roughly 2-hour development time disclosed by Karpathy is a personal project practice and does not mean that similar products can be completed at the same speed in terms of data processing, long-term maintenance, and security assurance. (Author/Yu Lei)

For more first-hand news, welcome to download the Phoenix News app and subscribe to Phoenix Tech. For in-depth reports, please search "Phoenix Tech" on WeChat.