6 min read
B06

Builders · No. 11

The Ten-Person Company

Instagram sold for a billion dollars with 13 people. AI is about to make that look crowded.

· · 6 min · 1,465 words

In short

  • Instagram had 13 employees when Facebook bought it for about a billion dollars in 2012, and WhatsApp had around 55 when Facebook agreed to buy it for 19 billion in 2014.
  • The number of communication lines in a team grows roughly with the square of its size, so ten people share 45 lines and a hundred share 4,950.
  • Software, cloud infrastructure and app stores turned distribution and servers into rented utilities, which is what made tiny giants possible.
  • AI extends the same leverage from infrastructure to the work itself, which shrinks the ideal team again.
  • In a very small company the bottleneck becomes taste and judgment about what to build, because execution is no longer scarce.

In April 2012, Facebook agreed to buy a photo app for about a billion dollars. The app was Instagram. It was roughly eighteen months old, it made no revenue, and it had 13 employees.

Two years later Facebook did it again, on a scale that still sounds like a typo. It agreed to buy WhatsApp for about 19 billion dollars. WhatsApp had hundreds of millions of people using it every month, all over the world. It had around 55 employees.

Those two deals did more than make a few people rich. They broke an assumption that had sat quietly underneath business for a century: that big outcomes need big organisations.

They do not. And the gap between the size of a team and the size of what it can do is about to get much wider.

Why small used to mean small

For most of history, output was tied to people. If you wanted to make twice as many cars, you needed roughly twice as many workers, twice the factory floor, twice the managers to supervise them. A company's headcount was a decent proxy for its power.

Even early software companies carried a lot of weight. You had to buy your own servers, rack them, cool them, and hire people to keep them alive at three in the morning. You had to ship discs in boxes to shops. You needed salespeople to get into those shops.

Then, over about a decade, most of that turned into a utility you could rent. Cloud computing meant a tiny team could run servers for millions of people without owning a single machine. App stores meant a teenager could reach every iPhone on Earth. Open-source software meant the hard parts of the stack were already written and free.

Instagram rode exactly that wave. Its founders, Kevin Systrom and Mike Krieger, did not build a data centre. They built a fast, simple app with good filters and a feed, ran it on rented infrastructure, and let the App Store do the distributing.

WhatsApp was even more extreme. The founders, Jan Koum and Brian Acton, were famously stubborn about simplicity. No ads, no games, no gimmicks: just messages that arrived reliably on almost any phone. They built on Erlang, a language designed decades earlier for telephone switches, which happened to be very good at handling enormous numbers of simultaneous connections. A few dozen engineers kept a global messaging network running.

They broke an assumption that had sat underneath business for a century: that big outcomes need big organisations.

The hidden tax on every new hire

Leverage explains how a small team can do a lot. It does not explain why small teams so often beat big ones. For that you need a bit of arithmetic.

In a team of n people, the number of possible one-to-one conversations is n times (n minus 1), divided by two. You do not need to love maths to feel what that formula does.

  • 5 people: 10 lines of communication.
  • 10 people: 45.
  • 50 people: 1,225.
  • 100 people: 4,950.

Headcount grows in a straight line. The web of conversations grows roughly with the square. Every new person adds hands, but they also add meetings, updates, misunderstandings, approvals and politics. At some point the coordination cost eats the extra output.

Fred Brooks saw this in the 1960s while managing a huge IBM software project. His 1975 book The Mythical Man-Month made the observation now called Brooks's law: adding people to a late software project makes it later. New people need to be trained by the people who were doing the work, and every addition multiplies the talking.

Jeff Bezos reached a similar conclusion from the other direction with his rule that a team should be small enough to feed with two pizzas. Pizza was beside the point. What mattered was keeping the web of conversation small enough that people could still move fast.

A ten-person company is a different kind of animal from a hundred-person company, rather than a cheaper copy of one. Everyone knows what everyone else is doing. Decisions take a conversation instead of a calendar. Nobody's job is to manage other people's jobs.

What small teams give up

It would be dishonest to pretend small is always better.

Small teams are fragile. When one person leaves, a fifth of the company's knowledge walks out the door. They can burn out, because there is nobody to hand the pager to. Some businesses genuinely need scale in people: a hospital, an airline, a bank with branches and regulators and call centres. You cannot run Emirates with thirteen people.

The real lesson from Instagram and WhatsApp is narrower and more useful. Headcount should follow the work, not precede it. Many companies hire ahead of need because hiring feels like progress. A bigger team looks like a more serious company. But each hire is a permanent increase in coordination cost, and you pay that tax forever.

The best small teams resist that instinct for as long as they can. They ask "what would we stop doing?" before they ask "who should we hire?"

A few habits show up again and again in teams that stay small on purpose:

  • They automate the second time, not the tenth. If a task comes back, someone writes a script, a template or a checklist before it becomes a job title.
  • They buy instead of build for anything that is not the core product. Payments, email, hosting and analytics are rented, not reinvented.
  • They write things down. A short written decision replaces a meeting that would have pulled in six people.
  • They say no to customers whose demands would bend the product into something else. Every special case is a future hire.
  • They hire generalists early who are comfortable owning a whole problem instead of one slice of it.

AI moves the leverage from servers to work

The cloud made infrastructure rentable. AI is starting to make a large share of the work itself rentable.

Ten years ago a small startup still needed separate people for many jobs. Someone to write the marketing copy. Someone to answer support emails. Someone to build internal tools, translate the app into Arabic, draft contracts, analyse the data, write the tests. Each of those was a hire, and each hire added lines to the web.

Today a capable builder with good AI tools can do a meaningful first version of most of those tasks alone. Not perfectly, and not without review, but well enough that the job no longer needs a full-time person in the early years. One engineer with strong tools can now ship what used to take a small team. A founder can test a dozen marketing angles in an afternoon.

That changes the shape of the ideal company. Sam Altman and others in the industry have talked openly about the idea of a billion-dollar company run by a single person. Whether or not that particular milestone happens soon, the direction is clear. Ten people with good leverage can now cover ground that recently required a hundred.

When execution gets cheap, judgment gets expensive

There is a catch, and it is the most important part of this whole story.

When making things becomes easy, the hard part moves. A small team with AI can build almost anything. So the question is no longer "can we build it?" It becomes "should we build it, and how good does it need to be?"

That is a question of taste and judgment, and it does not scale with tools. Erlang helped WhatsApp, but its real edge was a small group of people who were ruthless about what to leave out. Instagram began life as a more complicated check-in app called Burbn, and its founders cut it down to the one thing people loved, which was sharing photos. No server bought them that decision.

In a hundred-person company, a bad idea gets diluted by process, and a good idea gets diluted too. In a ten-person company, every choice goes straight into the product. That makes small teams terrifyingly honest. You see your judgment reflected back at you within weeks.

So if you are building now, stop asking how many people you can afford. Ask how few you can get away with, and whether each of them is someone whose judgment you would trust with the whole company.

The scarce thing in a small company is never hands. It is people who know what is worth doing.

A century ago, a great business needed a great factory. Then it needed great software. Increasingly it needs something far rarer and much harder to hire for: a handful of people who see clearly, choose well, and can each do the work of ten.

That is the ten-person company. It is not a small company. It is a big company with nowhere for anyone to hide.

Further reading

  • The Mythical Man-Month — Fred Brooks
  • Rework — Jason Fried and David Heinemeier Hansson
  • Company of One — Paul Jarvis

More from the shelf

B07 · Builders · 6 minDefault AliveOne question decides whether your startup survives, and most founders have never actually run the numbers. B08 · Builders · 6 minDo Things That Don't ScaleThe founders of Airbnb went door to door with a camera. The lesson was never about photos. N06 · Mind · 7 minThe Map Is Not the TerritoryEvery metric, résumé and org chart is a map. The trouble starts the day you forget it.

See the whole library →