Ask how it will fail before it ships

Working on systems serving more than 250,000 active users changed my definition of a ‘small’ engineering decision.

At scale, a slow query is not just a slow query. It becomes infrastructure cost. A vague error message becomes a support burden. A missing retry strategy becomes lost trust.

I learned to ask earlier: How will this fail? What will we measure? Can we roll it back? What happens when traffic is ten times higher?

Scale does not always require a complicated architecture. It requires deliberate decisions, useful observability and respect for the operational life of the code after deployment.

That is the mindset I now bring even to zero-to-one products.