I really like instruction counts for measuring performance. It's not perfect, but it's a very stable metric, even if I'm listening to music whilst developing and benchmarking.
miniblog.
Related Posts
I don't care what your project's test coverage is, but the fact that you're measuring it is a great sign.
On the challenge of measuring spam prevalence externally, because historical data tends to be cleaner:
Measuring Go program (GoAWK) performance from 1.2 to 1.18: https://benhoyt.com/writings/go-version-performance/
The Go 1.3 runtime led to a 2x speedup in GoAWK, and performance wins have continued since. It's easy to see popular PLs as largely finished, but the wins here show there's still opportunities.