Not Yet, if ever

technical

1. Distributed Tracing is basically best-effort distributed transactions, one to do the work, one to log
2. Sensible default, but configurable
3. Kubernetes because google buys a big enough hardware for only one app
4. “Lots of features” is good for services, “do one large thing (business functionality) well” is good for libraries
5. Is “infra as code” a code? If yes, how to test it?
6. Code to write VS to learn, or why big framework/abstraction is bad
7. Isolation in ACID is a very, very insanely expensive illusion
8. Async RPC framework is (mostly) useless, if not harmful
9. SLAs are kinda weird (better violating or just downtime?), which better be split into soft SLA and hard SLA
10. Automate for sake of automating?
11. There is always a place for middle value (memory vs NVM vs SSD vs HDD, GC + pool vs pure GC vs pure manual, all memory vs buffer manager vs wisckey, electron/react native/flutter, etc)
12. Tree structure as a model of concurrency among business entities
13. Simplicity by using the simplest silver bullet

non-technical

1. Stupid tech decisions and politics opening lots of new companies and job markets, and should be hyped even more
2. Machine is cheap, dev is expensive mantra only apply to countries with broken high salary