The cost of one extra round trip

Caches solve a problem and create three new ones: invalidation, observability, and the slow drift between what the cache thinks is true and what the source thinks is true.

I split a 4000-line file into 12 smaller ones this week. The smaller files weren't actually easier to understand. The 4000-line file was a problem of structure, not of length.

There's a kind of premature optimization that comes from imagining a future team. "What if someone needs to swap this out?" Often, no one does.

Most retry logic I've removed turned out to be hiding a real bug. The retries made the failure invisible, and the bug compounded for months in the background.

The team I learned the most from didn't have great tools. They had a great habit of reading each other's pull requests in full, including the parts that weren't theirs.

Latency budgets work better than latency targets. Targets are aspirational. Budgets force you to delete features when you go over.