There is a small group of people making genuinely excellent explanatory video about technical subjects, and this channel belongs in it. The style is distinctive. Long single-topic videos, heavy use of animation that carries the actual argument rather than sitting behind it as decoration, and a willingness to slow down and stay with a single idea until it is properly clear rather than covering ground. The result feels much closer to a well made documentary than to a tutorial, and it works on subjects that are usually presented either as formulas or as hand waving.
What makes it worth your time specifically, rather than being one more good YouTube channel, is the historical approach. A lot of this material is built around how an idea came about. What people were trying to do, what did not work, what constraint forced the shape of the eventual solution. That framing does something that formula-first teaching cannot.
When you learn a technique as a fact, you can apply it and you cannot reason about it. When you learn it as a solution to a problem someone had, you understand what it is for, which means you can tell when it is the wrong tool. The videos on the history of the field are the best example of this. Watching how the pieces came together, including the periods when the whole approach was considered a dead end, gives you a much better sense of why the current architecture looks the way it does than any amount of architecture diagrams.
The visual work deserves specific praise because it is doing real explanatory labour. There is a common failure mode in educational video where the animation is essentially illustration, pretty motion that accompanies a spoken explanation without adding to it. Here the visuals often are the explanation, showing a transformation or a space or a process in a way that makes the point before the narration arrives at it. That is much harder to make and much more effective.
Now the limitations, and they matter for how you use this. The catalogue is small. This is one person making things that clearly take an enormous amount of work, and the consequence is a slow publishing rate and a modest total volume. You can watch everything relevant in a weekend, which is a compliment to the quality and a real constraint on the value.
Coverage is idiosyncratic. Videos exist on subjects the maker found interesting, which is exactly how it should be for this kind of work and means there is no curriculum. There are gaps everywhere, and no path through the material, and no sense in which finishing it leaves you having covered anything. Treat it as a set of excellent essays rather than as a course.
The bigger risk is one that applies to all high quality explanatory video and is worth being explicit about. Watching a very good explanation feels like learning. It is pleasant, it is comprehensible, and you finish with the sense that you understand the thing. But there is no code, no exercises, and nothing to check your understanding against, and understanding an explanation is not the same as being able to produce the thing explained.
I have watched people come out of a run of excellent videos convinced they understood backpropagation, and then be unable to implement it. The videos were not at fault. The mode of learning has a ceiling and you have to go past it deliberately. So the way to use this is as a companion.
When a course introduces something and you can follow the mechanics but cannot see why it is that way, come here, watch the relevant video, and go back with the intuition. Used like that it is enormously valuable. Used on its own it produces a comfortable and slightly false sense of competence. My four point five is for explanatory craft at the very top of what is being made in this space, with a historical framing that teaches understanding rather than facts, and visual work that genuinely explains.
Marked down for a small and slowly growing catalogue, for no practice component at all, and for coverage that is a collection rather than a curriculum. Watch all of it, and then go and write some code.