Earlier quoted context omitted.
Hah, "just" a reinvention of graph analysis with linear-algebra. That's a "paradigm shift." Check out the performance numbers from that youtube video for redis-graph versus other graph databases. Reformulation of graph algorithms as algebraic operations means a huge step up in abstraction, and optimization, and performance. GraphBLAS is what enables that. It's the difference between tensorflow and having to rewrite s…
I'm not asking the point of graphs, I'm asking what is actually new. Directed acyclical graphs have been around for ages, combining them with linear algebra is not new. Instead of actually explaining anything that is new, original, unique or just a general step forward, all the replies seem to be saying how silly it is for asking and how smart someone is for counting triangles.
Re: Log(Graph): A Near-Optimal High-Performance Graph Representation (2018) [pdf]
#11Please take any single one of the sentences of my first paragraph above, to explain what's cool about the comment and the OP. Any one sentence would, if true, make this material far more worthy of hacker interest than the typical HN front page article. That the post above shows all of them makes this comment one of the better contributions to the site in the past week. I'm still making my way through all the material and without a doubt consider it the best information I've learned all week.