just dans

Making programming more accessible

02 Mar 2021

What will you do differently?

We’re in the fun stage at Replit. Usage and revenue keep growing week-over-week. The team remains small and tight-knit. If you want to build something, you can just build it and launch it. No one has the time to sit around telling people not to do things.

I’d love to stay in this stage for many years. The people we interview from larger companies all say the same thing: it was fun back when I started but now it’s not as fun and I can’t possibly know everyone.

We’re definitely a breath of fresh air for these folks. We only meet when necessary! Amjad & Haya have consistently let revenue grow well ahead of hiring. We won’t double in three months and then wonder why we can’t get anything done!

Yet if we follow the same trajectory as other growing startups, we’ll eventually get the same results. Then a smaller, growing startup can come along and nab all the bored people who wistfully remember the good old days.

What will we do differently? The two concrete things I can think of that we’re already doing differently are

  1. using Replit for internal tools whenever possible and
  2. giving each project a clear owner.

I’d love to add that we’re genuinely cool and kind people who care a lot about giving computing superpowers to the world. Our mission does set us apart. That plus the fact that the mission and the business are aligned - if we keep growing revenue we can keep giving away free compute to the world. We can have a great mission but still turn into a place that’s not fun to work at, though, so let’s focus on the two concrete day-to-day things we’re doing differently than other companies.

Building Replit means bootstrapping the new world out of the old. The difference between the two worlds becomes clear the moment you switch from working on an internal tool in a repl, where you can see what everyone’s doing and immediately run your code, to shepherding pull requests through review, CI, and deployment. Coding in a repl is more immediate and more fun than the traditional flow. The more tools we put into repls, the bigger the difference between being an engineer at Replit and an engineer at other companies.

Candidates often ask us how work happens at Replit. Are we top-down or bottom-up? Most projects grow out of conversations. Eventually someone gets fired up enough to write a design doc and demo a quick hack or some wireframes. That person now owns the project! They do a lot of the work, ask for help when they need it, roll it out, gather feedback from real humans, and blog about the project when it launches.

That all sounds pretty straightforward. What goes against the grain is that we have committed to strong ownership early. And we mean it. When problems arise, we choose the path that maintains strong ownership. If a job is going undone, it needs an owner. If a project has no end in sight, the owner needs to pick a tangible goal, write down what stands between them and that goal, and share their progress. It’s always tempting to do what’s expedient, to have veteran engineers step in and just fix the problem. But that would mean project owners are only owners on paper.

Those are the two things we’re doing differently right now. Both strategies will come under pressure as we grow the team. Especially as we approach the threshold where folks can no longer know the name of every single person at the company. We can delay the day we cross that threshold or decide to never cross it (though in that case we’ll trade the problems of scale for the problems of a relatively small group of insiders). It sounds ludicrous to think about that moment now, when the entire company still fits on grid view. But if we blindly cross that line we can expect to perform about as well as all the other companies with hundreds or thousands of employees.