Twitterment

Guide 06 / 06

A short history of searching Twitter, from Summize to X

Twitter's search began as a sentiment startup called Summize. Here is how it grew from there.

By the Twitterment Team · Updated · 4 min read

Searching Twitter is so ordinary now that it's easy to forget it wasn't part of the original product. In Twitter's first two years you could read your timeline and people's profiles, and that was about it. Finding what strangers were saying about a topic took a separate company, and that company started out measuring sentiment.

Summize: a sentiment engine that pivoted

Summize was founded around 2006 by former AOL employees. Its chief executive was Jay Virdy and its chief scientist was Abdur Chowdhury. Its first product had nothing to do with Twitter. It read product reviews and blog posts, worked out which words tended to go with high or low star ratings, and used them to score opinions in text that had no rating at all. TechCrunch covered it in December 2007 and called it a sentiment engine. Each product got a colored bar running from red to green.

During 2008 the team pointed the same technology at Twitter. Tweets were short, public and constant, and nobody else was summarizing them well. Summize's Twitter search let you type a word and see a live stream of posts containing it. It caught on quickly. Twitter itself worked with Summize in June 2008 during Apple's iPhone 3G keynote, a period when heavy traffic often knocked Twitter offline.

The acquisition in July 2008

On July 15, 2008, Twitter co-founder Evan Williams confirmed that Twitter had bought Summize. The price was not disclosed, and reports said it was paid mostly in stock. Twitter hired five of Summize's six employees, all engineers, which was a big addition for a company that had a dozen engineers of its own at the time. Summize's chief technology officer, Greg Pass, became Twitter's director of engineering and operations. Chowdhury went on to become Twitter's chief scientist.

The Summize search page became search.twitter.com. A search box later moved into Twitter's own interface, and the Summize name disappeared.

It's a neat bit of history for a site like this one. Twitter's search grew out of a company built to tell whether people liked something.

Search made a new kind of feature possible. If you can count which words appear in posts right now, you can see which ones are suddenly appearing much more than usual. Chowdhury built an early algorithm for this, watching for spikes in mentions of people, places and things. That work fed into Trends, the list of topics surging on the platform.

Hashtags had appeared before search did. In August 2007 Chris Messina suggested using the pound sign to group posts on a topic, and users adopted it. Search made hashtags far more useful, because clicking one could show you everyone using it.

Real-time search on Google and Bing

In October 2009 both Microsoft and Google announced deals to bring tweets into their search engines. Google launched real-time search in December 2009, showing a scrolling box of fresh posts from Twitter and other sources for topical searches. For a while, a news event would show tweets in Google results within moments.

That arrangement ended on July 2, 2011, when the agreement between Google and Twitter expired. Google took its real-time search offline the same week. Tweets kept appearing in Google through ordinary web crawling, but not as a live feed.

In 2015 the two companies made a new deal. Starting in May of that year, Google showed tweets prominently in its results again, this time with direct access to Twitter's stream of posts.

The complete index of 2014

For most of its early life, Twitter's own search only went back a short way. The original index was kept in memory for speed, which limited it to roughly the last week of posts. Older tweets existed but you couldn't search for them.

That changed in November 2014. Twitter's engineering team announced that it had built an index of every public tweet since 2006, close to half a trillion posts, stored on solid-state drives instead of memory. Queries still returned in a fraction of a second. Search results gained an "All" view that could reach back to the first posts on the service, and old tweets started turning up in searches again, sometimes to their authors' embarrassment.

Search on X today

Twitter was renamed X in 2023, and search now lives at x.com. The basics would be familiar to a 2008 user: type a word, see who is saying it. The Advanced Search form and the search operators layered on top, such as from:, since: and lang:, give much finer control. The guide on collecting posts by hand explains how to use them.

Sentiment, the thing Summize was built to measure, never became a built-in feature. Twitter's search shows you what people said and leaves the mood for you to judge. That is the gap the checker on the home page tries to fill for small batches of posts you collect yourself.

Questions

Who built Twitter search?
Twitter's search came from Summize, a startup founded by former AOL staff including chief executive Jay Virdy and chief scientist Abdur Chowdhury. Twitter bought it in July 2008. Chowdhury later became Twitter's chief scientist and built an early version of Trends.
When could you first search every old tweet?
In November 2014. Before that, Twitter's search index held only recent posts, roughly the last week or so. The new index covered every public tweet back to 2006.

Twitterment is independent and not affiliated with X Corp. Features and search operators on X change from time to time; if something here doesn't match what you see, check X's own help pages.

Open the sentiment checker