The future-proof startup

All of the production clusters I've retired have been ready to scale. As I pulled their plugs and listened to their disks spin down, as I slid them from their racks and thanked them for their service, each and every one of them was programmed to handle floods of users who never came.

When I terminated the production instances of my crisis management as a service company, those servers could dynamically scale to 1M active users per day.

I've written load balancers where there was no load. I've prepped hot failovers when there was no user to notice a cold failover. I've backed up offsite when the only user data was my team's. It was time which should have been spent growing the core product and user acquisition tools.

Before the A round, the only scaling preparation a web startup needs to worry about is picking materials that have been proven to scale. Before that time the team is too small, the workload too varied, and the user base too dinky to spend more than a day or two on scaling problems.

I can do a lot in a day or two: AWS now provides server instances powerful enough to get startups through the first year of business. Deployment tools like Chef come with massive libraries of administration scripts which, when combined with one of the many monitoring systems, give us the ability to quickly change configurations should we get a mention on Good Morning America. The days when we had to provision using a two year timeline are gone.

After the A round closes there is a moment when everyone realizes that something needs to be rewritten in order to scale for the next growth spurt. If you've been around the block a few times then it won't be something huge. But it often is and it feels bad. You think that you should have seen it coming so it feels bad.

Do you know what feels worse? Retiring your production machines and shutting down the business because you spent most of your dev time on scaling.

More posts

Recent posts

  • Trustworthy Augmented Reality Glasses

    Can we possibly trust augmented reality glasses?

    Below you will find a checklist for determining whether the device that you wear on your face is worthy of your trust.

    Each category has an introduction explaining ...

  • PotassiumES Dev Log #4

    Develop wider web apps for three display modes (flat, portal, and immersive) using just your desktop browser.

  • Cornerstone Organizations

    As we watch growth companies flourish and then disappear when acquired or sold for parts it seems useful to have a name for companies that actively choose to make tools that are durable and anti-fragile.

  • PotassiumES Devlog #3

    Using Three.js to build a reusable border geometry for spatial UIs!

  • PotassiumES Devlog #2

    🌸 Updated potassium style system (KSS), now with margins!
    🌸 The path to the vNext
    🌸 A couple of new spatial controls

  • PotassiumES Devlog #1

    🌸 A brief intro to the existing samples
    🌸 Building UI components that work in flat, portal, and immersive display modes

  • What is PotassiumES?

    Update: This is still a handy reference but you might be interested in the new PotassiumES site.


    This is a post about PotassiumES, an ECMAScript library that enables browser-side development for the wider web. If you're not sure about the wider web, click that ...

  • Wider Web Lingo

    Update: This is still a handy reference but you might be interested in the new wider web section of the PotassiumES site.


    People sling around a lot of lingo when talking about the wider web, and even the term "wider web" is lingo!

  • Wider Web Lingo: Voice

    There's a lot of lingo around the wider web so this is one of a series of short definition posts.

    Voice: Phrases or other vocal noises that can be recognized and used as input

    Computers are getting pretty good at understanding ...

  • Wider Web Lingo: Gesture

    There's a lot of lingo around the wider web so this is one of a series of short definition posts.

    Gesture: A body motion that can be recognized and used as input

    Computers are getting better at watching how we position ...