HackerTube Indexes 6,454 Hacker News Video Recommendations From 2007 to 2026
HackerTube, a collection of 6,454 YouTube videos recommended on Hacker News, covers recommendations from 2007 through 2026 and is organised across 150 topics. The details appear in a post dated 31 October 2026 on karpathy.bearblog.dev.
Engineers will notice the date range first. Recommendations from 2007 to 2026 cover nearly two decades of video links shared on one of the most widely read technical forums on the web. The 150 topics give the collection a structure that a raw list of links would lack, so a reader can move from a subject to the videos the community pointed to within it.
The format matters as much as the count. Much of the video that practitioners learn from is conference talks, live coding sessions, lectures and recorded demonstrations. These tend to be long, unevenly titled and scattered across channels, and they are hard to find again once the original feed has moved on. An index built on community recommendation addresses that problem directly, because the selection reflects a judgment about worth rather than a search query.
In my view, the more useful feature of a collection like this is the time axis. A recommendation is a timestamped act: someone with working knowledge of a subject judged a talk or demo worth sharing on a given day. Accumulated over nineteen years, those judgments form a record of which technical subjects drew attention at which points. Memory and search rankings rarely preserve that history well, and a dated archive of it can be read in ways a feed cannot.
The caution belongs with the signal itself. A recommendation records interest, and interest is not the same as correctness or currency. A widely shared talk may be dated, partly wrong or superseded by later work, so anyone using the archive as a reference should check its claims against current material. Popularity on a forum is a useful filter for finding things worth watching, but it is a weak guarantee of quality.
The broader context here is a change in how technical knowledge circulates. For much of the commercial web's history, practitioners learned from whatever a colleague forwarded, a bookmark they remembered, or a link that surfaced on a busy day. A curated, searchable index of community-selected video lowers the cost of rediscovering that material. The benefit extends beyond people who already follow the forum: a new engineer can start from the topics a large technical audience has repeatedly valued instead of from a blank search box.
The constraint is that an index is only as useful as the care taken in maintaining it. Topic assignment, removing duplicate entries and keeping links working all determine whether a collection like this remains a reliable reference in five years. Those are practical questions, and they will decide whether the archive becomes a lasting resource or a snapshot.
For engineers who already read Hacker News, the collection offers a different way in: a topic-indexed view of years of shared video, rather than the latest links on the front page. For everyone else, it is a compact map of what one technical community chose to recommend, and when. That is a modest thing to build, and a useful one, and it is the kind of tooling that tends to matter more over time than its launch suggests.
