from the leaked email said: "
“Hi Randy, after some internal discussion, here’s our preliminary list.

1) Intermittent glitching (”mech, intermit”) done in a way that’s more random sounding vs periodic.
2) Bit-resample, such that there is audible artifacting (sounds like a bad mp3 encode).
3) shifting channels (sounds like a speaker cut out). Again, the goal should be to sound somewhat random.
4) Laugh-track, at a respectable volume level.
5) Saw-tooth volume, so long as the volume goes to (or close to) zero, so that the track can’t be fixed by an inverse saw increase.
6) Beep, at a high volume

In the future, you might do experiments with static noise overlays (sounds like faulty recording equipment), voice over (public domain audio), and overlapping songs.

You probably don’t want to apply any effect for the first 30-60 seconds, so the user thinks they got a good track. We should take some care to ensure that when there is intermittent effects they happen in the same places so that it’s not possible to take the good portions of one version and splice them with the good portions of another version to get a complete (and perfect) third version.”"