I love reading the latest blog posts from the Slack Engineering team, Will Larsen and the Stripe crew, and the fingers-crossed-occasional post-mortem on painful outages. They are a treasure both to observe how these teams are operating and because freely sharing knowledge and experience is rare today. I am jealous to have the experience of service migrations under the load of tens of thousands of daily active users, or even millions. It creates a whole new level of challenge.
But that is not where me and my team are today. Not yet at least. There is no doubt that having these teams as a guide post, as a north star to start moving towards. It is invaluable to have a list of known possible issues next to possible solutions to evaluate. No, we are still searching for market fit. That means we have different engineering challenges.
We have all heard of the seed-round startup asking themselves "Should we use k8s or not?" and chuckling to ourselves, knowing that its the wrong question. It shows that they don't know what is important yet.
Engineering in the time of Product Market fit is about being able to present something to your potential customers As Fast As Possible. There is one truth—No One Knows Anything. Your organization does not know what customers will buy. Having the answer to that is called Product Market Fit. You don't have it yet.
I've learned that the best mindset for working in this environment is to expect that you're going to delete, shelve, and entirely ignore most of what you're building right now. Because no one wants it. This is a tough concept for some engineers who treat their code like the precious ring. As the saying goes today for servers: cattle, not pets. So it goes for your code and your prototype designs, learn to let go.
Time is a finite resource and the two largest constraints on your time is money and attention. You have a limited amount of runway. You only have so many attempts at gaining your customers attention. You only get to make a first impression to a new customer once. The more and more missed attempts and your potential customers will start ignoring you because they think they know what you're about.
The challenge today is to find the sweet spot. Just enough tooling to make development just easy enough to turn around MVPs to put in front of your customers to test. Because some of your tooling will end up being ignored and unused. The most important thing is getting the MVP out ASAP. The more time you spend on tooling, the less time you're spending on the actual MVP. The more time you spend on future proofing the less time you're spending on MVP. Don't optimize, don't scale. Until the moment you have to. Make plans. Don't act on them. It is a test of willpower and focus.