You know it's not going to be totally "from scratch." It's just the idea of how much did he build.
I mean if you wanted to be super hardcore, you could 3D print your own switches, your own key caps (or carve them out of wood), but what happens when you get to the controller? We don't have 3d printers for PCBs yet .. well none that are affordable for home use .. or that don't release tons of deadly toxins. :-P
Ironically, the PCB would probably take less effort than doing all of the switches and key caps yourself.
A simple matrix keyboard (downside: will misdetect certain keyboard combinations) requires a quite small set of io pins (if you insist on fully individual switches it'll complicate matters a bit, but not that much) that you could make a controller for with any number of simple old parts that are big enough that etching your own PCB for them at home is fairly straightforward.
For that matter good old 8-bit era parts like 8526/6526 CIA's (IO) and a 6510/8510 CPU are big enough and resilient enough (and still easy to get hold of) that you can build a controller on a breadboard, or even mount them on cardboard you punch holes in (... yes, been there, done that), and you can drive a serial port off them (in fact, the Amiga 500 and Amiga 2000 keyboards at the very least used a SOC version of the 6502 with onboard PROM and RAM for their keyboard controller)
Of course, going the "next step down" and building it all from transistors starts getting a bit more hairy.
For that "next step down" dig up Don Lancaster's "TV Typewriter Cookbook" for the 1976 version of this (and to understand "shift" and "control" at a more visceral level :-)
I mean if you wanted to be super hardcore, you could 3D print your own switches, your own key caps (or carve them out of wood), but what happens when you get to the controller? We don't have 3d printers for PCBs yet .. well none that are affordable for home use .. or that don't release tons of deadly toxins. :-P