Pretty sure it was the way hi-res graphics worked on the VIC-20 too -- 22 characters by 8 pixels per character = 176 pixels wide, multi-colour mode where each block of 8x8 pixels had a background and a foreground colour. Deeply freaky.

The Vic also had a mode whereby you could make "double height" characters (halving the number of addressable rows), so when you typed "a", it would emit (iirc):

  a
  b

and when you typed "b", it would emit:

  c
  d

When I was about 8, I fooled with this peeking and poking pixels so that "a" and "b" bitmaps where the top and bottom halves of a double-height "a", through to the end of the alphabet. I imagined that this would make it easy for my grandparents to use computers. Never mind that they couldn't type, had no interest, and would never remember commands like "load $,8,1".

edit: formatting

On the ZX81 and ZX Spectrum, they called them "User-Defined Graphics" or UDGs. I think I wrote a pixel-flipping grid font editor for characters on every computer I had access to! The ZX81, the ZX Spectrum, the Apple II (once I noticed that one of the games I had (Taipan?) hooked the text output routines into a hires character generator), and eventually various x86 programming environments, although there you were editing bitmaps, not changing the character generator source. :-)

I used semigraphics on a CoCoII, not my video, but gives an idea of what that means ;)

https://www.youtube.com/watch?v=F2rscrAGCaM

> You used to be able to define your own characters on Amstrads

ah so THAT is how you could make graphics on those machines? wish I had known that back when I was 10 :-P

ah well, good times :)