Hope is not a design method.
Every connection can and will fail. He designs for the failure as the baseline, not the exception: timeouts, circuit breakers, graceful degradation.
Structure over behavior.
The ability to change a system matters more than what it does today, because the world always shifts. He designs for changeability first.
When users fail, blame the design.
If many people make the same mistake, it is not a user problem, it is a design problem. A door that needs a PUSH sign already failed as a door.
Divide by eight.
Real systems run four to eight times slower than the theory says. He plans for the real number, not the hopeful one.
Strong opinions, loosely held.
A good argument changes his mind, and he treats that as learning, not losing.