100twt.com

History

A table that reloaded every three minutes

What 100twt.com was, how it worked, what it caught during 2009, and why nothing like it could run the same way now.

What it was

100twt.com went up in early May 2009 with a single question as its headline: what if you were subscribed to the 100 most popular people on Twitter? Below the question sat one long table. Each row was a tweet, newest first, from one of the hundred most-followed accounts. A meta refresh reloaded the page every three minutes, so leaving it open in a tab turned it into a slow ticker.

Each row had four parts: the account's avatar; the handle and its most recent tweet; the time it was posted, linked to a discussion page; and the follower count, with the account's rank underneath, linked to TwitterCounter. A later version added a one-click RT link that opened Twitter with the tweet ready to retweet.

The discussion pages

Every tweet that entered the river got a numbered page of its own, holding the tweet and a comment thread hosted by Disqus. The counter tells you how fast it ran. The page numbered 00009 is dated May 4, 2009; 00957 is dated May 6; the first homepage capture, on May 9, is already linking to numbers around 02790. More on the discussion pages.

Timeline from the captures

  1. May 4, 2009Discussion page 00009 is created. The app is running.
  2. May 9, 2009First capture of the homepage. 37 accounts in the 100 newest tweets; @aplusk ranked #1.
  3. May 15–19, 2009The Twitter staff river is running with a daily archive, and the New York Times river is captured. The nav's Blog link points to a post dated May 16 about following the people of the Times.
  4. May 20, 2009The Gillmor Gang river is captured.
  5. June 4 and July 12, 2009Two more homepage captures. The July page carries the RT column and a nav bar: TOP-100, Twitter, NYT, Gang, Blog.
  6. September 2009The Berkeley and Twin Cities rivers appear, each linking to a post on another site that introduced it.
  7. November 2, 2009The latest timestamp the homepage ever shows: 7:27 PM Pacific.
  8. January 4 and March 12, 2010The archive saves the homepage twice more. Both copies are byte-for-byte the same and still show the November 2 table.

The race to a million

The app arrived just after the most public contest in Twitter's early history. In April 2009 the actor Ashton Kutcher challenged CNN's breaking-news account to reach a million followers first, and won. The captures pick up the aftermath. @aplusk ranked first whenever he appeared on the table, with 1,660,243 followers on May 9 and 3,897,128 on November 2. In July, @cnnbrk was third with 2,248,254. Further down the table, @petewentz was still counting toward his own million.

The bottom of the list says as much about 2009 as the top. In May the account ranked #100 had 352,358 followers. Across the captures, bloggers, a productivity author, Twitter's own founders, a cat (@sockington) and a deals site all sat within the hundred. The network was still small enough for that.

What happened to the API

100twt worked because Twitter's 2009 API was open in a way it never was again. A script could read public timelines and follower counts with very little setup, which is why one person could build a river and then a handful of sibling rivers in a few months. Twitter retired that first version of its API in June 2013 and required authenticated, rate-limited access in its place. In 2023, under new ownership, it moved most API access to paid tiers.

The homepage stopped updating long before any of that. After the app went quiet the domain passed through other hands and was used for unrelated material, none of which is reproduced here. No copy of the app is running anywhere else; it depended on an API that no longer exists.

How this site was built

Everything on the data pages comes from the Internet Archive's captures of 100twt.com and its subdomains. The homepage table was saved on May 9, June 4 and July 13, 2009 (showing July 12) and on January 4 and March 12, 2010 (both showing November 2, 2009). Each table was parsed from the raw HTML: handle, tweet, time, follower count and rank. Nothing was interpolated. If an account didn't tweet inside a capture's window it has no reading for that date, and the table shows a dash.

Tweets are quoted sparingly and exactly as the table printed them. Avatars aren't reproduced.

Credit

The original 100twt.com (2009) was an experiment by Dave Winer. This site is an independent retrospective and is not affiliated with him. The original table's styling was credited on the page to Veerle Pieters' CSS table design.