Stewart Butterfield's Two Failed Games Became Flickr and Slack

On a gray November morning in 2012, Stewart Butterfield walked along the seawall in Vancouver’s Yaletown with one of his employees and told him the game was over. Glitch, the online world his company Tiny Speck had spent three years and most of its money building, was going to close.

“If we keep going as we are, we’ll burn through the rest of our money in a few months and be left with nothing to show for it,” Butterfield said, according to Johnny Rodgers, who was there and later wrote about it. “But if we stop now, we can use that money to build something else. I think the tools we’ve built internally could be useful to other people.”

Those tools became Slack. Eight years later, Salesforce agreed to buy the company for $27.7 billion.

What makes the story unusual is that it had happened to Butterfield once before.

The first game

In the summer of 2002, Butterfield co-founded a small company in Vancouver called Ludicorp with Caterina Fake and Jason Classon. Its project was Game Neverending, a multiplayer online game. It never caught on, and after a couple of years it was shelved.

The team had built a photo-sharing tool as part of the game, and decided it could stand on its own. They built it out as a separate product, and in early 2004 launched it as Flickr. It arrived at exactly the moment blogging was taking off, and it grew to tens of millions of users. In March 2005 Yahoo bought Ludicorp. The price was never officially disclosed; it has been widely reported at around $25 million.

Butterfield stayed at Yahoo to run Flickr and grew frustrated with life inside a large and distracted company. When his vesting period ended in 2008 he left. His resignation letter said he would spend more time “tending to my small but growing alpaca herd.”

The second game

He hadn’t given up on games. In 2009 he co-founded Tiny Speck with a group that included Cal Henderson, Flickr’s lead engineer, as well as Eric Costello and Serguei Mourachov. Its project was Glitch, a whimsical browser game set inside the imaginations of sleeping giants. It launched on September 27, 2011.

Glitch had a loyal community, but not enough players to pay for itself. Rodgers lists what the team tried: alpha and beta releases, new features, a rebuilt experience for new players, invitation campaigns, credits for existing players. None of it worked well enough. And the world was changing under them. Glitch was built in Flash, which didn’t run on smartphones, just as casual gaming moved to phones.

By the autumn of 2012 the company had about 40 people. Butterfield faced a choice many founders get wrong. He could keep spending in the hope of a turnaround, or he could stop while there was still money and goodwill left. He stopped.

The team announced the closure in November. Players were refunded for recent purchases. The company set up a page to help its staff find new jobs and later released much of the game’s art for public use. Glitch’s servers went dark on December 9, 2012, at 8 p.m., after a farewell party attended by thousands of players.

If we stop now, we can use that money to build something else.

Stewart Butterfield, November 2012, as recalled by Johnny Rodgers

A company without email

The tools Butterfield mentioned on the seawall were not a product. They were a pile of workarounds. Tiny Speck had staff in Vancouver, San Francisco and New York, and it ran almost entirely without email. Instead, according to Rodgers, everyone lived in a chat system built on IRC, an old internet chat protocol, with channels for different topics: #general, #deploys, #files, a busy channel for support.

Over time the engineers bolted on extras. Files uploaded to a server appeared in a channel. Chat logs went into a database so they could be searched. Sign-ups, purchases, support requests, code deployments, app store reviews and tweets were all piped into channels automatically. It was text-only and held together with scripts. But it meant a spread-out team always knew what was going on.

Butterfield’s pitch to the few people he asked to stay was to rebuild that system properly, as a single searchable place for a team’s communication, connected to the other software it used. He suggested the name Slack. The investors who had backed Glitch agreed to let the company use the remaining money for the new direction. Within a month, the team was using a first prototype.

Slack takes off

Slack was announced in August 2013 and offered first to a small group of companies. It launched publicly in February 2014. Growth was immediate: Wired reported weekly growth of 5% to 10% that year, and by the first week of August 2014 more than 120,000 people were using it daily.

Much of that came from word of mouth, as users recommended it to colleagues and friends. By December 2015, according to The Wall Street Journal, Slack had raised $340 million from venture investors and had more than 2 million daily active users and about 570,000 paying customers.

Butterfield’s approach to selling it was set out in an internal memo in 2013 that he later published under the title “We Don’t Sell Saddles Here.” The argument, roughly, was that people don’t buy a product for its features. They buy what it does for them. Slack wasn’t selling channels and integrations; it was selling teams less email, less chaos and a better working day. The memo has since become a widely shared piece of startup writing.

The exit

Slack went public on the New York Stock Exchange in June 2019, through a direct listing rather than a traditional IPO. Its shares opened at $38.50, giving it a market value above $20 billion. Then Microsoft, which bundled its rival Teams with Office, turned the market into a fight Slack found hard to win on its own.

In December 2020 Salesforce agreed to buy Slack Technologies for $27.7 billion, and the deal closed in July 2021. Butterfield stayed on to run Slack inside Salesforce until he announced his departure in December 2022. He left in early January 2023.

The rule of the second try

Butterfield’s career is easy to tell as a story about lucky accidents. Twice he set out to build a game, and twice something else came out of it. That reading misses what he actually did each time.

He paid attention to what people used. At Ludicorp the players loved the photo feature more than the game. At Tiny Speck his own staff couldn’t work without their chat system. In both cases the evidence was right in front of him.

He stopped before the money ran out. Plenty of founders spend their last dollar on a failing idea out of pride or hope. Butterfield’s decision on the seawall meant Tiny Speck still had cash, staff and investors when it started on Slack. A failure that ends early leaves assets behind. One that ends late leaves only debts.

And he kept the team. The people who built Glitch were the people who built Slack, and they already knew how to work together. That helps explain how a working prototype was ready within a month.

It’s a pattern with relatives elsewhere in this archive. Bubble Wrap failed twice before IBM gave it a job, and Arm found its future only after its first big customer flopped.