More concretely learn how to write specs in Gherkin/Cucumber and back them with TLA+ and Alloy before you code. Remember the speed of light (about 1 ns per foot) and annotate those specs with best case data travel latency.

Use holistic profiling like https://twitter.com/OptimyzeCloud/status/1351203905850519557

When you see performance issues then dig into your language stack and learn how to remediate them.