HackerTube: A YouTube Built from HN Comment Recommendations, Featuring 6,454 Videos

A Video Discovery Engine Driven by HN Comments

Hacker News generates a massive volume of comments every day, and hidden within them are countless high-quality video recommendations. But the timeline nature of comment sections means these recommendations quickly sink and become hard for later readers to discover. HackerTube attempts to solve this problem: it extracts YouTube videos mentioned in HN comments over the past two decades and reorganizes them with a YouTube-style interface, letting users browse content curated by the HN community just like scrolling through a video site.

According to the project's launch post on HN, HackerTube currently contains 6,454 videos, sourced from 39,834 HN comments, spanning from 2007 to the present. The launch post itself received 284 points and 115 comments, indicating that the community has considerable interest in this idea of "using HN data to feed video consumption."

Where the Data Comes From and How It's Filtered

The core of HackerTube is not algorithmic recommendation but rather the reuse of human recommendation signals. Specifically, it identifies YouTube links in HN comments and then aggregates those links into a playable video library.

The Value of Comment Sections as Recommendation Signals

HN's comment culture has a distinctive trait: users only post links when they believe a video is genuinely worth watching. This amounts to an initial screening done manually by community members, filtering out a large amount of low-quality content. Compared with YouTube's own recommendation algorithm, this signal is closer to "peer review"—the recommender must stake their own reputation on it.

Hacker News's ranking mechanism itself also reinforces the quality of recommendation signals in comments. HN uses a decay algorithm based on votes and time (the Gravity algorithm); highly upvoted comments briefly rise to the top but are quickly buried by new content. This "rapid rotation" design encourages users to share resources with genuine long-term value in comments rather than chasing trends. At the same time, HN's real-name/pseudonymous culture lets users accumulate reputation (karma), and recommending low-quality content may damage personal credibility, thereby forming an informal community self-governance filter.

Content Depth Brought by the Time Span

The time span from 2007 to the present means that HackerTube's library contains early internet tech talks, classic conference talks, and recent developer interviews all at once. This vertical depth is hard for ordinary video recommendation feeds to provide, and it is better suited to veteran readers who want to systematically mine technical content.

From 2007 to the present, the HN community's technical focus has shifted several times: early on it was mainly startups, programming languages, and hardware hacks; in the 2010s it shifted toward distributed systems and machine learning; and in recent years it has focused on AI safety, large models, and the open-source ecosystem. HackerTube's video library thus forms a chronological curve of technical interests, and users can intuitively perceive the evolution of tech hotspots by browsing the most frequently mentioned videos from different years.

Product Experience: Familiar UI, Different Content Logic

HackerTube adopts a YouTube-style interface, meaning users do not need to learn a new interaction model. Elements such as the thumbnail grid, playback page, and related recommendations all follow common video site conventions, lowering the barrier to use.

But beneath this "familiarity" lies a difference in content logic. The homepage of an ordinary video platform is driven by click-through rate and watch time, whereas HackerTube's content ordering essentially reflects the distribution of attention within the HN community—which videos are mentioned repeatedly, and which topics cluster in particular years, will all be reflected in the browsing experience. For users accustomed to the HN context, this is more like "a visualization interface for community memory."

The core value of a YouTube-style interface lies in reducing cognitive load. Users can quickly judge content relevance through thumbnails, titles, duration, and view counts (or HN mention counts) without learning a new information architecture. But HackerTube cannot rely on YouTube's personalized recommendation algorithm, so its "related recommendations" are more likely based on co-occurrence relationships in HN comments or video tags, which requires the system to perform additional semantic analysis or clustering of video content.

The Demand This Kind of Project Reflects

HackerTube's emergence is not an isolated phenomenon. It points to a genuinely existing pain point: high-quality recommendations are scattered across discussion threads and lack a searchable, browsable aggregated form. HN's search function can find comments containing specific keywords, but it cannot organize content by "video" as a unit, nor can it provide a visual browsing experience.

Extracting 6,454 videos from 39,834 comments—this ratio shows that the density of video recommendations in comments is not low, but the vast majority have never been systematically organized. What HackerTube does is turn these scattered signals into structured assets.

Implications for Content Creators

If a video can keep appearing in HN comments, it has long-term reference value. Such content is often not hot-trend-driven fast-moving consumer goods but rather tutorials, technical analyses, or industry talks. For creators making in-depth content, citation in HN comments can serve as an observational metric—it reflects recognition from a peer group rather than exposure from algorithmic distribution.

Limitations and Open Questions

HackerTube currently only handles YouTube links; video recommendations from other platforms are not covered. In addition, the rules for deduplication and ranking, and whether dead links are included, are not explained in detail in the original launch information. The 115 HN comments very likely contain discussion of these details, but this cannot be confirmed from the launch post alone.

For users who want to try it, they can visit hackertube.org directly to browse. As an experiment in reusing community data, its value lies not only in providing a usable video library but also in demonstrating an approach: turning implicit recommendation signals in discussion areas into an interactive product form.

Handling dead links is a common challenge for aggregation projects. YouTube videos may become unavailable due to copyright issues, deletion, or account bans, and HackerTube needs to regularly verify link status and flag invalid content. In addition, deduplication must account for the same video being mentioned multiple times (possibly in different comments); a reasonable approach is to use the video ID as the unique identifier and record the comment time and context of each mention, so as to support later ranking or display of community discussion热度.