Books, and what stayed with me

Takeaway

Error budgets changed how I argue with clients. Perfect uptime is not the goal, agreeing on how much unreliability is acceptable is, and once that number exists the technical decisions stop being opinions.

Takeaway

The slow feedback loop is always the dangerous one. A site that degrades over eighteen months never triggers an alarm, and nobody is thanked for fixing something that had not broken yet.

Takeaway

Fix the broken window immediately. Most of the sites I inherit are not badly built, they are well built with two years of small unfixed things stacked on top.

Takeaway

Reading the chapter on replication lag explained three separate bugs I had previously blamed on caching. Knowing the storage layer is not optional, even when you did not choose it.

Takeaway

Experts skip steps under pressure, which is exactly when steps matter. My incident checklist is short, dull and has saved me from more mistakes than any tool I own.

Takeaway

The identity part landed harder than the habit part. I stopped saying I should train and started being someone who trains, and the same trick works for the boring maintenance nobody feels like doing on a Tuesday.

Takeaway

Useful on protecting attention, less useful for anyone whose job is interruption. I took the mornings and left the rest.