well... im only 18, yet I remember messing around with my brother's TI99/4A.

that thing was sweet, as a matter of fact, I still have it in a closet, and it works perfectly except for a broken RF Adapter. It seems ironic now, I bet most digital watches have more power, with its AMAZING 16k of ram, the 16-bit TMS-9900 processor running at a zippy whole 3Mhz and would display at a nice resolution of 252x192, catch this, IN COLOR!!! (okay, maybe that was ahead of its time)

Most over europe the CoCo was sold as "Dragon32" or "Dragon64"
chears
A friend of mine had a dragon
Yes, and it was the first that was realy fun to program in Assembler..
.. i assembled on a peace of paper used a table to search the hex values and entered it with a hex editor .. no kidding ..
just small tools like a asci-print for the "HighRes" Graphic Display..
An addaption of the "Amiga Ball" and other small things..
The bigest thing was a selfmade Scanner which was based on a reflexLED and some resistors, which was mounted on the printer head and connected with the analog-Joystick port..
Very funny, very cheap, but it works..
..i had to reprogram the joyport-routin because the buildin was not fast enough...

I know you're not kidding, because I did it exactly the same thing
The first machine code program I wrote was actually on the HeathKit 6800 CPU trainer. I remember writing a digital clock program, completely in machine code, and it was about ~300 bytes (or less)
The company I used to work for (circa 1985, for this issue), had a 20 channel VHF/UHF IMTS telephone switch terminal for car telephones, and the main board was a Motorola 6809 processor with 32KB of RAM.
I had fun changing the 1 second interval NMI timer vector, and pointing it to my own machine code routine, for doing weird things like generating test tones, capturing phone lines, forcing manual connection between a phone line and a channel, etc.
I had a single one byte value, that if changed, it would execute my own vectored routine followed by the original (saved on stack), and then clear the byte. Otherwise, the regular NMI routine would be called. It was fun, and also fun crashing the system when using incorrect OP codes
But system reset was only 5 seconds, even with a 2Mhz clock
I always liked Motorola's CPU arquitectures better than Intel's

-kwag