Developer

How Andrej Karpathy Built HackerTube with Claude Code and Opus 5.5

Built by Andrej Karpathy · · Source: X

Quick answer

Andrej Karpathy had Claude filter 43M Hacker News comments from 20 years down to 6,454 videos that keep getting recommended, then build and deploy HackerTube, a video site with no logins.

Result: According to Karpathy, a significantly better viewing experience than YouTube's recommendations, built in about 2 hours with Opus 5.5.

What they built

Andrej Karpathy was annoyed with YouTube's recommendations, so they spent a weekend building HackerTube (hackertube.org). According to Karpathy, Claude filtered 43M comments from 20 years of Hacker News down to 6,454 videos that keep being recommended there. The site shows those videos as its recommendations.

How it works

Karpathy says they use Claude Code (the desktop app for Mac) with Opus 5.5 on Medium effort as their default. They only talk to Claude and have it do everything. The post lists three things they did by hand in the browser:

  • Created an API key in the Google Cloud UI so Claude could query YouTube for video metadata
  • Booted up a VPS box to serve the site
  • Bought the domain name

Once the box was up, Karpathy gave Claude the IP and told it to "update, secure, harden, deploy and optimize the code". They say it did the whole thing.

The post describes these features of the site:

  • No logins. State is kept in the local browser session.
  • Plain thumbnails. Each one is a representative frame from the middle of the video instead of the uploader's A/B tested image.
  • Open data. The underlying data from the comment analysis is available as one zip file, and Karpathy suggests pointing your own Claude at it and asking for your own version.
  • HackerTube Longs. The post lists this by name without explaining it.

The result

Karpathy calls it a significantly better viewing experience and says it took about 2 hours with Opus 5.5. They add that many projects that would normally have taken weeks or months can now be done very quickly.

How to build something similar

  • Pick a community whose taste you trust and find a public archive of its posts or comments.
  • Ask your coding agent to extract every link of the type you care about (videos, books, papers) and rank them by how often they come up.
  • Do the account steps yourself, such as API keys, a server and a domain, then hand the agent the details and let it build and deploy.
  • Keep the first version small: a static list with state saved in the browser is enough to see if the recommendations are any good.

The original post

Weekend hack: HackerTube. I was annoyed with YouTube's recommendations so I had Claude filter down 43M comments over 20 years of HN to 6,454 videos that keep being recommended. *Significantly* better viewing experience, ~2 hours with Opus 5.5: https://hackertube.org/ - Excellent recs without any logins (state is kept in local browser session) - Thumbnail is a representative frame from middle of the video instead of the super duper A/B test hyper-optimized user uploads - Underlying data from the comment analysis as one zip file, point your Claude at it and ask for your own version - HackerTube Longs ✊ A bit more on the process: I use Claude Code (desktop app for Mac), Opus 5.5 on Medium effort as my default. I only talk to Claude and have it do everything, the only "manual work" I helped with in tandem was in the browser (reminiscent of MenuGen): 1) I had to create some kind of API key from Google Cloud UI to query YouTube for video metadata, 2) I booted up a VPS box for serving and 3) I bought the domain name. Once the box was up I gave Claude the IP and told it update, secure, harden, deploy and optimize the code, and it did the whole thing. Many awesome projects that would have normally taken weeks/months even can now be few-shotted for breakfast. What else are we not thinking of.
Andrej Karpathy (@karpathy) · November 1, 2026 · View on X

Want to build something like this?

Join the Shipping Skool community free. Paid members get live bootcamp calls every week.

Join Shipping Skool — Free

More use cases

← All AI agent use cases