Consider that Aphex's Windowlicker was a created at around the same time as the sound the author made, and it was an entirely computer sequenced track (with lots of effects programmed in the SuperCollider audio programming language):

http://www.youtube.com/watch?v=iZ8sZXFN6jc

That should give you some idea of the audio effects possible on computers at the time. The Tri-Tone is child's play by comparison.

You notice he went with a fairly limited set of notes; if you're not a musician take my word for it that these are notes guaranteed to sound good together, because a) they're a succession of whole-number frequency ratios and b) they are also in the ratio of the first few Fibonacci numbers, and the human brain really, really likes hearing those basic tones recombined.

All the stuff with LISP & perl seems like a complete and utter waste of time to me. Even the author says he was 'geeking out' - presumably to avoid putting his musical ego on the line.

It would have been vastly simpler and easier to open up a MIDI sequencer (choosing from one of the many available on the market, and almost every musician has at least one installed), and either play in a bunch of chords and then offset the notes, or gone into the event list editor, tossed in the first few combinations, and then cut-n-pasted the rest.

There isn't anything wrong with doing it the way he did - it seems like he wanted to push his personal musical taste to the background, and use a deliberative approach to combine and play the notes without any possible musical bias. I have algorithmic musical software that I sometimes use for the same reason, especially when I'm trying to learn something or break out of a musical habit. But there were many less time-consuming ways of achieving the same result.

But there was no need to do things this way in 1998. Even then it would have looked like using a sledgehammer to crack a nut.