The YouTube videos are better because they need it to work on screen.

Videos are slow, and it's hard to re watch tricky pieces or compare with other sources.

I wish someone in the ML community would take this on: create a document from video, maybe using the appropriate frames of the video as media in the document. Give people something to glance through instead of having to watch a whole video. Bonus points for making the document interactive with links to appropriate sections of the video as well as links out to the world for products/definitions/etc. Double-bonus points for the ability to squelch awful humor and filler words.

Until Google decides to take down one for this or that reason. I would rather use a personal blog (backed up at archive.org just in case) with all the necessary text information, plus links to a video hosted on a decentralized video hosting platform, then maybe publish the video, or part of it, on YouTube as well to attract audience.

To play videos you need a working video driver and a working soundcard.