The Outage You Couldn’t Sleep Through

This week AWS East went down and the internet lost its mind.

Twitter was full of redundancy and failover complaints. Redditors argued about how poorly everything’s architected. Even senators jumped in to say Amazon’s gotta get broken up.

Fine debates to have. But if you’re someone actually building something new, you might have missed the real lesson.

8 Sleep went down with it #

I’ve been trying to get my sleep in order lately. Sleep casts. Meditation. Breathe in for 4, hold 7, out for 8. Magnesium. Whatever Ashwagandha is?

And yes, I even got one of those bougie mattress pads — the 8 Sleep Pod. It’s this smart cover that heats, cools, and tracks your sleep with the kind of data you’d expect from NASA.

I thought I was maybe figuring this sleep thing out when two nights ago, I woke up to a 47 sleep score. Out of 100. Grade F.

But it wasn’t because I had a restless night — it was because the Pod thought I’d been awake since 3 a.m. when it couldn’t talk to the cloud anymore. AWS East was down.

My NASA-level smart pad became a flat-earther.

And I wasn’t alone — threads popped up full of people saying their Pods weren’t working. Some guy (with the even more ludicrously priced Pod that reclines up and down) said his version wouldn’t recline anymore. So he had to sleep on the floor.

The 8 Sleep founder wasn’t ignoring this. He posted that their team was staying up all night to make sure the rest of us could… not stay up all night.

The thing is — it didn’t really matter #

Yeah, customers were annoyed. Me included. But I didn’t throw the Pod out. I didn’t cancel my subscription.

Because: most people didn’t care that much. They knew it’d be fixed in a few hours.

And look at 8 Sleep. They’ve gotten this far — multiple generations of hardware, glowing press, a half-billion-dollar valuation — with a single point of failure sleeping (I swear I made this awful pun up myself) right in the middle of their stack.

Sure, maybe someone should have improved their redundancy before version 5, hundreds of thousands of customers, and a Series D of $100 million.

But the real lesson is that they ignored the AWS multi-region debate for years — and it didn’t stop them from building an incredible product and company. And I doubt there’s much fallout for them now. Besides needing to read their own blog about How To Make Up For a Bad Night of Sleep.

Startups don’t die because of AWS outages #

Every time AWS sneezes, engineers rush to Medium to write about why you must design for cross-region redundancy and draw diagrams about how you can do it too.

But if you’re a startup, you probably shouldn’t.

Because the thing that’s going to kill your company isn’t downtime — it’s nobody caring. It’s that your product doesn’t solve anything meaningful yet.

You’ve got so many other problems to figure out first:

– finding even a single user
– getting them to pay
– shipping something fast enough to learn why they won’t

Christ, you probably have to worry about paying cloud taxes before you need multi region failover headaches.

Of course, not every business can ignore redundancy. If you’re running ventilators, air traffic control, or a stock exchange — yeah, go ahead and worry.

But for 90% of people building something new?

You can’t optimize for a future where you’re 8 Sleep until you’ve survived long enough to be 8 Sleep.