Somewhere in a Vancouver office in late 2012, a company was quietly dying. Tiny Speck had spent nearly four years and roughly $17 million building an online world set inside the minds of eleven sleeping giants, and almost nobody wanted to live there.
The game was called Glitch, and you have probably never heard of it. What the team built to talk to each other while making it became one of the most recognizable products in modern office software. The wreckage turned out to be worth more than the ship.
The game was supposed to be the whole company
Eleven giants, forty employees, and one very expensive dream
Tiny Speck was founded in 2009 in Vancouver by Stewart Butterfield, along with Cal Henderson, Eric Costello, and Serguei Mourachov, all Flickr veterans. The company existed to build a browser-based massively multiplayer game conceived as an alternative to combat-heavy titles. This was not a side bet or a hedge. The game was the entire reason the company drew breath.
Glitch was strange in a way that is difficult to convey secondhand. The world was set inside the heads of eleven giants “who thought of funny things until their thinking came alive,” and players took the role of beings inside those thoughts. It deliberately avoided combat mechanics, focusing instead on collaborative crafting and gathering, with players growing plants and trees, cooking, and reshaping a persistent world together. Skills carried names like Bubble Tuning and Bureaucratic Arts. Some skills took real-world hours to learn, and the whole thing emphasized exploration and socializing over the grind. Keita Takahashi, the designer behind Katamari Damacy, was among the people working on it.
Butterfield had a theory of what he was building. On launch day, he explained the distinction between finite games, which have fixed rules and are played to win and thereby end, and infinite games, which have no fixed boundaries and whose purpose is simply to continue. Glitch, he said, was an infinite game. He was pointed about the contrast with Zynga, arguing that Glitch was deeper and more open-ended, built that way because the team thought it was beautiful rather than because it would monetize better.
Investors bought the vision. Tiny Speck raised a $5 million Series A from Accel and Andreessen Horowitz, followed by a $10.7 million Series B in April 2011, bringing the total to $17.2 million since its founding. The company grew to roughly 40 artists, engineers, game designers, animators, musicians, writers, and in-game guides.
There is a detail here worth holding onto. Butterfield began conceiving the game that would become Glitch as early as 2002, while at Ludicorp. Financial pressure pushed the team toward a side project instead, and that side project became Flickr. Glitch was not a first attempt. It was a second run at an idea he had never stopped believing in.
Preposterous, beautiful, and mostly empty
Beloved by thousands, which was precisely the problem
Glitch officially launched on September 27, 2011. It lasted about two months in that state.
On November 30, 2011, Tiny Speck “unlaunched” the game and reverted it to beta. The reasoning was candid. Butterfield wrote that two large improvements were needed. First, the early experience had to reveal itself more clearly, allowing new players to reach the enjoyable parts sooner. Second, players who had already fallen in love with Glitch needed stronger creative tools that would let them reshape the world in meaningful ways, including building new locations and forming organizations capable of larger collaborative projects.
The announcement acknowledged that making radical changes to the game’s core system was much harder while it remained publicly available. The team also needed time to improve stability and prepare the infrastructure for growth. Butterfield later explained that a formal unlaunch was necessary because Glitch was already generating revenue and could not simply be treated like an ordinary beta.
None of this sounds like a company that had lost faith in its game. It sounds like one that had identified the problem and was buying itself time to solve it. However, the fix simply did not arrive in time.
The trouble was structural rather than aesthetic. Glitch was free to play with in-game purchases, and that model only works at volume. Critical affection does not pay server bills, and a small, devoted community generates a fraction of what the economics require. And so, the closure announcement went out in mid-November 2012 and remains one of the more honest documents the games industry has produced. Tiny Speck wrote: “Glitch has not attracted an audience large enough to sustain itself, and that, based on a long period of experimentation, it seemed unlikely it ever would.”
Given the movement toward mobile and the decline of the Flash platform, it was unlikely to get there before its time ran out. The game, they said, had pushed the limits of what a browser-based game could do, and then those limits pushed back. Under an FAQ heading titled “Why why why why?” came the line that has outlived the game: they had set out to build something crazy, beautiful, and worthwhile that would succeed creatively and financially, but “we only managed to create a crazy, beautiful game.”
Tiny Speck had also tried to find a buyer willing to keep it running, and failed.
What happened next says something about the company. Butterfield set about finding new jobs for the roughly 30 people being laid off, winding down the forums, and issuing refunds to players still in credit. Internally, the task was called “Hire a Genius,” a page listing every employee’s photo, skills, and contact details, linked from every closure announcement and press release. The team released the game’s code, artwork, and music into the public domain, a decision that later allowed the Odd Giants project to reconstitute a playable version. The servers went dark on December 9, 2012, at 8 pm, after an End of the World party attended by thousands of players.
Buried in that same closure announcement was a sentence almost nobody noticed at the time. “We have developed some unique messaging technology with applications outside of the gaming world,” Butterfield wrote, “and a smaller core team will be working to develop new products.”
The useful part was hiding behind the game
Everything Tiny Speck built so it would never have to use email
To understand what he meant, you have to look at how Tiny Speck actually worked.
The team was split between offices in Vancouver and San Francisco, with at least one co-founder working remotely from New York. Johnny Rodgers, who joined in early 2012 and stayed through the company’s pivot, described the onboarding process bluntly: there was effectively no email. Most communication happened in Internet Relay Chat (IRC), the decades-old chat protocol. The company server had channels organized by purpose, including #general for company-wide conversation, #deploys for code releases, and #support-hose for incoming support requests.
That was not novel by itself. Plenty of engineering teams lived in IRC. What Tiny Speck did was keep bolting things onto it.
They added a simple FTP server that announced uploads into a #files channel. It wrote a system that stored chat logs in a database so conversations could be searched and old archives revisited. It also built what we now call Slack integrations that pushed updates from other systems into IRC, so new Glitch signups, credit purchases, support requests, code deployments, App Store reviews, and tweets from the company account all appeared inside the same workspace.
By the time the system matured, Tiny Speck could talk in real time, share files, search for almost anything anyone had said, and watch the business unfold without wading through an email thread. Rodgers was careful not to romanticize it. By his account, it was a system only a nerd could love: quirky, mostly without a graphical interface, requiring you to set up your own clients and bookmark separate pages for each subsystem, held together with custom scripts, cron jobs, and ugly SQL queries. It worked well enough because it only had to serve Tiny Speck. Every ounce of the team’s real attention was still going into Glitch.
The decisive conversation came on a rainy morning in November 2012. Butterfield took Rodgers on a walk along the Yaletown seawall and explained that Glitch had no viable business. Continuing would burn the remaining money and leave the company with nothing, but stopping now meant they could use what was left to build something else. “I think the tools we’ve built internally could be useful to other people,” he said. Rodgers asked, with some understandable skepticism, whether he meant the IRC server.
Butterfield’s pitch was already rehearsed: all of a team’s communication in one place, synced, searchable, and available wherever you go. One system replaced email and file hosting, and a single integration point for everything a team needed to say to itself, both between humans and between people and machines. They would throw out the duct tape and start fresh while keeping what they had learned.
He also already had a name. “I’m thinking of calling it Slack,” he said, making a gesture like loosening a string between his fingers. “We can come up with a better name later.” The four co-founders had made up their minds, and the investors were already on board.
Slack escaped the wreckage
Three lists, one month, and a board demo that landed
The eight people who remained at Tiny Speck started writing code in January 2013.
They began with three lists: Channels, for what conversations you can read; Messages, for what a conversation is about; and Members, for who is in it. The interesting work went into the details that made a text log feel human.
A read cursor tracked which messages you had seen and synced across web and mobile, so you could start reading on your commute and resume at your desk. Consecutive messages from the same author were grouped together, giving equal visual weight to people who typed in short bursts and those who preferred carefully composed paragraphs. Whenever a conversation paused for five minutes, Slack inserted a visual gap to make long stretches of real-time chat easier to scan. The team also began extracting links from messages and fetching their metadata. A bare URL could expand into what Slack called an “unfurl,” complete with a title, description, and image.
Then came the two features that mattered most. You could drag a file in, click to add it, or paste it from the clipboard. Links to Google Docs or Dropbox files were fetched and unfurled automatically. All of it was indexed alongside the surrounding message history, which meant a shared PDF could be found by searching the text inside the file, while a Google Doc could be retrieved by searching a phrase it contained.
That became the basis of Butterfield’s demonstration to the board in early February 2013, only a month after development began. He wrote a Google Doc in one window, shared its link in Slack, then searched Slack for a phrase from the document. The result returned both the file and the conversation in which it had been shared, anticipating the same ability to search inside Google Drive files shared through Slack, which later became a standard product feature. The board was impressed. The larger idea was that the many disconnected tools inside a workplace could be tied together by a searchable messaging layer where every event, file, and update could become a message.
On February 14, 2013, the team decided it was good enough and moved off IRC entirely. Butterfield announced it in typical fashion: “NO MORE WORK TALK ALLOWED IN IRC! $500 fine per incident.”
The codename nobody could kill
They spent months avoiding the name they’d already picked
Slack is widely reported to stand for Searchable Log of All Conversation and Knowledge, although the company’s own history leaves the origin pleasantly unresolved. According to Slack, the word first appeared as Stewart Butterfield’s codename for the project, with the title page of an early pitch deck reading “codename /slack” in the chunky slab typeface used for board presentations at the time.
The team tried repeatedly to replace the name and never found anything better. A trademark search eliminated Honeycomb, while Superset was already taken. Cedar, Pecan, Loop.in, Kitchen, and a joke option called chatly.io all came and went. Slack’s iOS app is still registered as com.tinyspeck.chatlyio, because the team had not chosen a final name when it needed to submit an early build to the App Store. By the time Tiny Speck acquired slack.com, thousands of people were already using the iOS app, and forcing them to download a replacement would have created unnecessary friction.
The domain belonged to a man in Montana whose personal website featured programming scripts and photographs of cats. Tiny Speck initially offered $8,000 and received no response. It raised the offer to $25,000, but it was turned down. The eventual deal closed at $60,000, and the first version of slack.com went live in early June 2013.
How early testers exposed Slack’s scaling flaws
Quit email for a week
External testing began before Slack was ready, and it exposed weaknesses almost immediately. Butterfield asked a handful of friendly startups to abandon email for a full week and conduct all of their internal communication inside Slack. His reasoning was straightforward. A partial rollout would only split conversations across multiple tools and hide the real problems.
Rdio was one of the first companies to try it. Until then, Slack had never been used by more than eight people at once, and the team had only guessed it might scale to 100, or perhaps 150, users. Rdio had more than 300 employees. Within days, it had spawned hundreds of channels and direct-message conversations, pushing the software into territory its creators had barely considered. That pressure produced features such as muting conversations, trimming stale direct-message lists, and much faster channel switching. Testers also invented naming conventions without being asked, naturally organizing discussions into channels such as #design-mobile, #design-desktop, #sales-america, and #sales-asia. The resulting sprawl was an early version of the problem users now solve by organizing the Slack sidebar. The product was beginning to learn from its users almost as quickly as its users were learning it.
That feedback loop is the part of Slack’s early history I find most revealing. A /feedback slash command let anyone report a bug without leaving the app, sending those reports directly into the team’s own #support-hose channel. It was common for someone to identify the issue, deploy a fix, and ask the original reporter to refresh the page, all within an hour. The cycle repeated dozens of times each day, and Butterfield himself frequently replied to customers.
Slack’s preview release came in August 2013. Butterfield has said it was essentially a beta, but they refused to call it one because people assume betas are flaky. With a press push built on the team’s existing reputation, 8,000 people requested invitations on the first day, and 15,000 within two weeks. They then spent six months working through those signups and incrementally improving the new-user experience before opening to everyone in February 2014.
What made it stick was not that it was chat. IRC was chat, and so were HipChat and Campfire. Slack’s difference was that it treated a work conversation as a durable, searchable, indexed archive that other software could write into, and then wrapped that in an interface a non-engineer would tolerate. Email had durability but no presence or integration surface. IRC had the immediacy, but nothing persisted unless you built the plumbing yourself, which is precisely what Tiny Speck had spent four years doing by hand.
Eight months to a billion dollars
This was a game studio nobody could buy
The numbers arrived fast enough to embarrass the game. By October 2014, Slack had raised $120 million at a $1.12 billion valuation, roughly eight months after its public launch. Butterfield was candid about the round number, noting that $1 billion is better than $800 million because it is the psychological threshold for customers, employees, and the press. At the time, 45,000 teams and 365,000 daily active users were paying. By January 2015, it had 500,000 daily active users.
On June 20, 2019, Slack Technologies went public through a direct listing on the NYSE under the ticker WORK, skipping underwriters entirely. The NYSE set a reference price of $26, and the stock opened at $38.50. It closed the first day at $38.62, valuing the company at about $19.5 billion, nearly three times its last private valuation of $7.1 billion. Seven years earlier, the same company had been unable to find anyone willing to take a video game off its hands for free.
Here’s where Salesforce stepped in
The lifeboat docks at Salesforce Tower
The ending was less triumphant than that suggests. Microsoft Teams arrived bundled with Office 365 at no additional cost, bringing several of the advantages that could make Microsoft Teams better than Slack into software that many businesses were already paying for. Competing with software people already have is a difficult argument to win. By late 2020, Slack’s position on the public market looked underwhelming. It had net losses of $147.6 million across the two quarters ending July 31, and, in TechCrunch’s assessment, it was a sitting target.
On December 1, 2020, Salesforce and Slack announced a definitive agreement under which Slack shareholders would receive $26.79 in cash and 0.0776 Salesforce shares per Slack share, for an enterprise value of approximately $27.7 billion. That announcement did not mean the acquisition had closed. The US Department of Justice requested additional information in February 2021, ended its investigation on July 16, and the transaction was finally completed on July 21, 2021.
Butterfield left in early 2023, less than two years after the acquisition. In his departure note, he said he was not planning another startup and wanted to spend more time with his family, gardening, and looking after his health. Slack has since cycled through Lidiane Jones and Denise Dresser. Dresser departed for OpenAI in December 2025 to become its first chief revenue officer, and Rob Seaman, previously chief product officer, took over as interim CEO. He still holds the role on that basis as of this writing.
Slack today is no longer really a chat company, which is the most Salesforce thing that could have happened to it. On January 13, 2026, Salesforce announced the general availability of a rebuilt Slackbot positioned as a personal agent for work, with Slack framed as the conversational interface where people and agents work together. It runs on Anthropic’s Claude. Furthermore, on March 31, 2026, Salesforce announced more than 30 additional capabilities,
Slack’s own operational figures indicate the platform handles more than 700 million messages per day, runs nearly 3 million workflows daily, and has approximately 1.7 million integrated applications used each week. These are company-supplied measurements rather than directly comparable market-share figures, but they show how far the system has moved beyond the eight people who abandoned IRC in February 2013.
- OS
-
Web browser, Windows, macOS, Linux, Android, and iOS.
- Developer
-
Slack Technologies, a Salesforce company.
- Price model
-
Freemium, with a permanent Free plan and paid Pro, Business+, and Enterprise+ subscriptions.
Slack organizes workplace conversations into channels, direct messages, huddles, and searchable shared spaces or teams. It also connects workplace apps, files, workflows, and AI tools within one collaborative platform.
The giants slept, but the messages survived
Slack survived because nobody at Tiny Speck had treated its communication system as precious. It was the accumulated residue of people solving practical annoyances one script, database table, and integration at a time. Its features had been shaped by real work rather than a theory about what offices might someday need.
Glitch’s world was dreamed by eleven giants and inhabited by too few humans. Slack endured because it recorded the work required to keep that world standing. The cathedral disappeared, but the scaffolding had already learned how to travel.