I believe the library mentioned in the article should be this one https://geodyssey.github.io/overview/overview.html and here https://www.lukatela.com/geodyssey/

The author says:

"It used Voronoi Cells and reduced most math to 8- and 16-bit integer calculations with one or two single-precision floating-point calculations."

I wonder if it can be used to speed up LLM inference.