No kidding. Didn't know that. I wonder who I first followed. - DeWitt Clinton

Ah, it was @miyagawa. - DeWitt Clinton

Cool huh :) - Dion Almaer

Now I'm wondering if the userids were really handed out sequentially and without gaps. Mine is 673483, @miyagawa's is 731253, @dalmaer's is 4216361. Could nearly 1 million people have signed up for Twitter before I did? Or 4 million people before you did? I created my account 2007-01-21, @miyagawa on 2007-01-30, and you on 2007-04-11. Were there really 3 million new accounts in 3 months? - DeWitt Clinton

In case you were curious, here's how I'm checking: http://json-indent.appspot.com/indent... - DeWitt Clinton

Mine is 817083 and I joined 2007-03-07 - EricaJoy

Hmm, so 3,399,278 in 35 days? - DeWitt Clinton

Wow. Tempting to create a new account and see what number you get :) - Dion Almaer

Interesting. I just started plugging in n+1 around your id, Erica. Accounts are sparse indeed there. For example: http://json-indent.appspot.com/indent... - DeWitt Clinton

I suspect that they allocated the ids in large chunks across sharded dbs, and never reclaimed the unassigned ids. - DeWitt Clinton

Whether or not my hypothesis is true about why it happened, I think we can easily demonstrate via random sampling across the range 0..current_max_id that the current_max_id is a poor indicator of the actual number of accounts. - DeWitt Clinton

Wrote the script. Very *very* interesting how sparse it is. Running a statistically meaningful sample now. - DeWitt Clinton

Yup, this is going to be good. The Twitter API is rate limited to 100 per hour, so I will need to trickle in data slowly. I'll let it run all night. Hmm. - DeWitt Clinton

Finished up the post sampling Twitter data. Will publish it tomorrow AM. Neat stuff. - DeWitt Clinton

Posted: http://blog.unto.net/twitter... - DeWitt Clinton