Reliability is a design decision.
A system is defined as much by the way it fails as by the work it performs.
Start with the boundary.
Every system operates within a set of assumptions. Network access, valid inputs, available capacity, and trustworthy participants may look like implementation details. In practice, they define the conditions under which a design can keep its promises. Reliability begins by making those conditions explicit.
Make failure legible.
A failure that is visible can be investigated. A failure that resembles success may persist unnoticed. Clear states, traceable decisions, and honest uncertainty give operators the information they need to respond. The objective is not the appearance of certainty; it is an accurate account of what the system knows.
Prefer a smaller promise.
A narrow guarantee that remains true under pressure can be more useful than broad capability with poorly understood limits. Define the promise, identify what could break it, and design the response before extending the scope.