TL;DR
Get business pricing on monitors, keyboards and dev gear
- Business-only prices and quantity discounts
- Tax-exempt purchasing
- Multiple users, one account, clear invoices
HackerTube is a YouTube-style site assembled from videos recommended in Hacker News comments. Its creator says an AI-assisted workflow filtered comments to identify 6,454 videos across 150 topics; the project uses local browser state rather than user accounts.
HackerTube, a YouTube-style video discovery site built from links recommended in Hacker News comments, was introduced by its creator in a report dated October 31, 2026. The creator says an AI-assisted process filtered the forum’s comment archive to identify 6,454 YouTube videos recommended from 2007 through 2026, offering an alternative way to browse videos without relying on YouTube’s standard recommendation feed.
In a post on his personal blog, the developer identified as karpathy said he started the project after finding YouTube’s recommendations unhelpful. He described the feed as showing exaggerated thumbnails and repetitive subjects, but that assessment is his account of his own experience, not an independent evaluation of YouTube’s recommendation system.
According to the post, an AI workflow processed Hacker News comments, filtered them, and used a model to help remove spam and lower-quality material. The resulting collection contains about 40,000 comments recommending videos, which the creator says yielded 6,454 distinct YouTube videos across 150 topics. The post does not provide a separate audit of the filtering or counting process.
The site presents the collection in a YouTube-like interface. The creator says it uses a thumbnail frame taken at the 50% point of a video rather than the uploaded thumbnail. He also describes custom “For You” recommendations, keyboard shortcuts, and the option to export site data for use in creating another version. The post says recommendation state remains in the user’s local browser session and that the site does not require accounts.
A Different Route to Video Discovery
HackerTube matters as an example of a small, purpose-built discovery service using community recommendations as its organizing principle. Instead of ranking videos around a viewer’s account history, it draws from links people shared in public Hacker News discussions. That could make it useful to visitors looking for material that has been recommended in conversations about technology and other subjects.
The project also shows how a developer can combine an existing online archive with AI-assisted filtering and a familiar interface. The creator says the process took roughly two hours, though that is a self-reported estimate and does not establish how much time would be needed to reproduce or maintain the service. The source does not compare HackerTube’s results with YouTube’s recommendations or establish that its selected videos are objectively higher quality.
For users, the stated lack of accounts and use of local browser state offer a different arrangement from services that personalize feeds through logged-in activity. But the post does not explain the full data practices of the site, how its recommendations are generated, or whether those practices have been independently reviewed.
Products Worth Considering
Bose QuietComfort Wireless Noise Cancelling Headphones, Sandstone
Logitech MX Mechanical Wireless Illuminated Performance Keyboard – Clicky
Logitech MX Keys S Wireless Keyboard, Graphite
As an affiliate, we earn on qualifying purchases.
From Hacker News Comments to Videos
Hacker News is a discussion site where users share links and comment on technology, science, business, and other topics. Recommendations in its comments are informal: users mention videos in conversations rather than submitting them through a dedicated review process. HackerTube’s creator says the project turns those scattered mentions into a browsable video collection.
The reported source period runs from 2007 to 2026. The post says the processing pipeline narrowed the material to 40,000 comments that recommended a YouTube video, producing a library of 6,454 videos. It does not spell out how repeated links were counted, how topics were assigned, or what rules distinguished a high-quality recommendation from a low-quality one.
In the original report, dated October 31, 2026, the creator described using an AI coding assistant and a model-assisted workflow, then packaging the collection in a YouTube-style interface and deploying it to a server. These are details reported by the creator; the post is not an independent technical review of the site or the tools used.
“Let’s write our own YouTube clone based on whatever videos have been recommended on Hacker News.”
— karpathy, HackerTube’s creator
How the Collection Was Filtered
The public account does not provide the filtering criteria, the prompts or steps used in the model-assisted workflow, or an independent check of the resulting collection. It is not clear how the project handles duplicate links, removed videos, age restrictions, or videos that have changed or become unavailable since users recommended them.
The creator says recommendations are customized without accounts and that state is kept in a local browser session, but gives limited detail on how the “For You” feed selects videos or what data the site may collect through other means. The post also does not report traffic, user feedback, or whether HackerTube will be maintained over time. Its description of YouTube’s feed as repetitive is the creator’s opinion, not a measured comparison.
Availability and Project Details
The creator’s report presents HackerTube as a deployed site and says visitors can export its data to help build their own version. No future release schedule, maintenance plan, or additional features are specified in the source material. Readers seeking to evaluate the project can check the site itself and the creator’s post for updates, while treating the reported collection size and build time as claims from the project’s author.
Further information that would clarify the service includes its full data and privacy practices, the exact method for generating recommendations, and how often the video list is refreshed. The source does not say whether those details will be published.
Key Questions
What is HackerTube?
HackerTube is a YouTube-style site that organizes videos mentioned in Hacker News comments. Its creator says the collection includes 6,454 videos across 150 topics.
Where did the videos come from?
The creator says an AI-assisted workflow processed Hacker News comments from 2007 through 2026, identifying about 40,000 comments that recommended a YouTube video. The post does not include an independent audit of the process.
Does HackerTube require an account?
According to its creator, the site does not require user accounts and keeps recommendation state in the local browser session. The source does not give a full account of the site’s data practices.
How are HackerTube recommendations generated?
The creator describes the site as offering custom “For You” recommendations but does not explain the ranking method in detail. The exact criteria and how the feed uses browser-session state remain unclear.
Has HackerTube been independently evaluated?
The supplied report is the creator’s account of the project. It does not include independent testing of the video selection, recommendation quality, privacy practices, or reported build time.
Source: karpathy.bearblog.dev
