HackerTube: YouTube Built From Videos Recommended By HN
AIThis post was created with the assistance of artificial intelligence (AI).

TL;DR

Before you orderOffer from Amazon

Get the latest gadgets delivered free with Prime

  • Fast, free delivery on millions of items
  • Prime Video, Amazon Music and more included
  • Member-only deals all year
Start your free Prime trial Free trial for eligible customers · Cancel anytime
As an affiliate, we earn on qualifying purchases.

A developer writing as karpathy says they built HackerTube, a YouTube-style site made from videos recommended in Hacker News comments. The site’s creator reports that it includes 6,454 videos across 150 topics, with recommendations assembled from comment data and browsing state kept in the user’s browser.

A developer writing as karpathy says they have published HackerTube, a YouTube-style site that organizes videos recommended in Hacker News comments. The project turns a reported collection of 6,454 videos mentioned in comments from 2007 to 2026 into a browsable feed, offering an alternative to YouTube’s usual recommendation interface.

In a post dated October 31, 2026, the creator described HackerTube as a response to a YouTube home page they found crowded with repetitive recommendations and exaggerated thumbnails. They said they asked an AI coding agent, identified in the post as Opus 5.5, to build a video site from recommendations made on Hacker News. That description and the site’s reported scale come from the creator; the supplied material does not include an independent audit of the data or process.

According to the post, the workflow collected Hacker News comments, filtered them, and used a second model, Sonnet, to help remove spam and lower-quality material. The creator says this produced about 40,000 comments recommending a YouTube video, from which the project identified 6,454 videos spanning 150 topics. The post does not explain how duplicate links, dead videos, or differing opinions about quality were handled.

The site is presented in a YouTube-like interface and includes topic browsing, a custom “For You” recommendation feed, keyboard shortcuts, and an option to export the site’s data. The creator says it does not require user accounts and keeps recommendation state in the browser session. They also say it uses the middle frame of a video as its thumbnail rather than YouTube’s uploaded thumbnail. No figures for visitors, viewing activity, or recommendation performance were provided.

At a glance
announcementWhen: Published October 31, 2026; the creator…
The developmentA developer has published HackerTube, a YouTube-style interface populated with videos recommended in Hacker News comments.

A Community-Sourced Video Feed

HackerTube offers a small-scale example of a different way to organize online video: instead of relying on a platform’s personalized history, it starts with links recommended in a technical discussion community. For viewers who find their usual feed repetitive, a collection formed from other people’s recommendations may surface videos they would not encounter through their existing YouTube habits.

The project also highlights how AI-assisted filtering and coding can be used to turn an existing discussion archive into a standalone interface. That does not establish that the resulting selections are more accurate, more diverse, or better than YouTube’s recommendations. It does show how a public archive can be repackaged into a focused discovery tool, while raising practical questions about selection criteria, stale links, and whose preferences the source community represents.

Because the creator says the site works without accounts and stores state locally, it also presents a different privacy arrangement from services that build recommendations around logged-in viewing histories. The source does not provide a technical review of data handling, so that feature should be understood as the creator’s description rather than an independently verified privacy assessment.

Products Worth Considering

Bose QuietComfort Headphones (2nd Gen), Wireless Noise-Cancelling Over-Ear Headphones, Black

Bose QuietComfort Headphones (2nd Gen), Wireless Noise-Cancelling Over-Ear Headphones, Black

Bose QuietComfort Wireless Noise Cancelling Headphones, Sandstone

Bose QuietComfort Wireless Noise Cancelling Headphones, Sandstone

LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor

LG UltraGear 32GX850A-B 32-Inch 4K UHD Glossy OLED Gaming Monitor

As an affiliate, we earn on qualifying purchases.

From HN Comments to Videos

Hacker News is a discussion site where users submit links and comment on them. Recommendations for videos can appear within discussions about programming, science, technology, and other subjects; HackerTube’s creator says the project gathered such recommendations from comments posted between 2007 and 2026. The project therefore uses a community archive as its source rather than claiming to index all YouTube videos or represent the preferences of YouTube viewers generally.

The creator’s account describes a short build process: collecting and filtering comments, extracting video recommendations, then placing the results in a familiar video-browsing layout. They estimate the work took about two hours, crediting Opus 5.5 for the coding task and describing Sonnet’s role in the filtering workflow. These are statements in the project post; no code review, dataset documentation, or independent timing estimate is included in the supplied reporting.

“Let’s write our own YouTube clone based on whatever videos have been recommended on Hacker News.”

— HackerTube creator karpathy

How the Collection Was Filtered

The available account does not provide a public description of the full filtering rules, the number of duplicate or unavailable links, or how the system classified a comment as a recommendation. It is also unclear how “quality” was defined when spam and lower-grade material were removed, and whether the 40,000 comments represent distinct users, videos, or recommendation events.

The creator’s reported counts and workflow have not been independently verified in the supplied material. There are no published measures of how well the “For You” feed performs, how often it updates, or whether the collection favors particular topics or frequent Hacker News contributors. The post also does not specify how long locally stored session data remains available or provide a separate technical audit of the site.

Site Access and Data Details

The creator’s post points readers to hackertube.org and says the site’s data can be exported for people who want to build their own version. The post does not give a roadmap, release schedule, or commitment to update the collection, so it is not clear whether the video list will grow or be refreshed as new Hacker News comments appear.

Further detail from the creator could clarify the extraction and filtering method, the treatment of broken or removed YouTube links, and how the recommendation feed is generated. Until those details are available, the headline counts and descriptions of the interface remain the project author’s account of HackerTube’s launch.

Key Questions

What is HackerTube?

HackerTube is a YouTube-style video browsing site populated with links that its creator says were recommended in Hacker News comments.

How many videos does it include?

The creator reports 6,454 videos across 150 topics, extracted from about 40,000 comments. Those figures come from the project post and have not been independently verified in the supplied material.

Does HackerTube require an account?

The creator says it works without user accounts and keeps recommendation state in the browser session. The available description does not include an independent technical assessment of its data handling.

How were the videos selected?

According to the creator, the project collected Hacker News comments, filtered them, and used AI models to help remove spam and lower-quality material before extracting video recommendations. The exact criteria and review process were not detailed.

Will the collection keep updating?

The post does not say whether HackerTube will regularly add new recommendations or refresh its existing links. No update schedule or product roadmap was provided.

Source: karpathy.bearblog.dev

You May Also Like

YouTube is expanding its AI deepfake detection tool to all adult users

YouTube is now expanding its AI likeness detection tool to all users over 18, allowing individuals to monitor and request removal of deepfake content using their likeness.

Pinterest Surges In Global Coverage

Pinterest’s media mentions have surged globally, with GDELT reporting 27 times the baseline in recent coverage, signaling rising interest and visibility.

Gta 6

Rockstar Games confirms development of GTA 6, with a planned release date in late 2024 or early 2025, sparking widespread excitement among gamers.

Tech Burnout: When Always-On Becomes Too Much

Growing dependence on constant connectivity can lead to tech burnout, but understanding how to manage your digital habits is essential to prevent…