How did you arrive at that conclusion? I assume Twitter retains everything as well (even "deleted" tweets) and it's all associated with an email address. Or did you mean it in the sense that far fewer people have a Twitter account?
These are tar files that contain bz2 compressed newline separated twitter events as json. These include deletion events as well, so you can for instance easily estimate the time an auto-deleter is set to.
Yes, they're huge archives, but you could still probably process a year of these for particular targets for under $10 on EC2.
Whilst I'm impressed with archive team's efforts, I would be surprised if there aren't some commercial twitter stream consumers that absolutely dwarf this.
Treat everything you put on twitter as public forever and you won't go too far wrong.
Because of the twitter stream APIs it's not. But there does seem to be a strange presumption amongst users that deleted tweets are gone from public view and cannot resurface. There are people who use tweets in all manner of ways that they really weren't designed for, some of which involve deleting them after a few minutes.
Many a public figure uses these tweet deletion apps. Some do it for more honest reasons (status count limits -- do they still exist?), others do it to limit their exposure.
In the UK at least, there have been cases of libel where either the claimant or the defendant depended upon twitter and in at least one of these the court admitted the claimant had an unfair advantage by forgetting about having a tweet deletion app attached to their account. The case proceeded and the claimant won despite the acknowledged advantage. To some, this may be seen as a clear message that in the eyes of the judiciary it's okay to delete tweets (evidence) as long as it was through an auto deletion app and the individual concerned forgot about its existence.
I would not at all be surprised if some lawyers to the rich quietly suggest they install a tweet deletion app as general advice upon instruction.