In response, and as an addition, to Wilbert Baans’ storytelling with public databases, I made a little twitter scraper. It indexed all twits for the query http://search.twitter.com/search?q=%22i+voted%22+%2B+McCain+OR+Obama+-twitvote, up till a month ago. From November the 4th till November the 5th the script scrapes new twits for this query every minute.

A point goes to Obama if the regular expression /vote.*?obama/i succeeds, it goes to McCain if the regular expression /vote.*mccain/i succeeds, else it is undecided / unrecognized.

The result can be found at Twitter Poll. So far Obama is winning.

Update: Wilbert Baan has made theses stats into a beautiful visualization:

There is an other twitter poll on Mashable.

Update 2: I tried making a similar exit poll based on MySpace, but that query does not yield results sorted on time. Also, just like Google, MySpace only returns a maximum of 1000 results per query. Unfortunately I’m thus stuck with the same numbers, no matter how many times I will scrape it.