At a basic level it's extruding the individual faces and then setting the position of those faces, scaling the edge vectors. I must be honest here - the exact math setup for this node I actually borrowed/found from a question somewhere on Blender Stack Exchange a long time ago, back when my understanding of Geometry Nodes was far lesser than it is now.
I've never bothered to properly unpick the exact workings of it because it just works, and I've been using it in my workflows ever since. However I've just spent the last hour or so working through the math and I think basically what it is doing is replicating the solve that I describe in the next lesson for offseting curves correctly by multiplying the values by the square root of 2 (i.e. the hypotenuse of a 1x1 right angled triangle)
There's also some additional stuff going on that keeps everything uniform if you have non-right angle edges, but that's the basics of it. So much of the time with Geometry Nodes, other people have already created and/or solved the exact problem that you face, there's no sense in re-inventing the wheel. It's perfectly fine to utilize other people's node setups, just like in coding it's absolutely fine and accepted to use snippets of other people's code found on Stack Overflow etc.