π in Julia is not just a Float64 constant. It's evaluated to the accuracy required of the type you're using! https://julialang.org/blog/2017/03/piday
miniblog.
Related Posts
An optimistic take on neural networks for programming: https://medium.com/@karpathy/software-2-0-a64152b37c35
It makes some good points about predictable runtime performance, ability to trade CPU for accuracy, and the ease of hardware acceleration.
When you're writing static analysis tools, should you make your analysis flow or path dependent? What is the accuracy/performance tradeoff?
https://www.youtube.com/watch?v=JpK9e__q5Ts shows an elegant approach using monad transformers to make this pluggable.
Fun post on CSS minification tricks: https://luisant.ca/css-opts-survey2 (shows the huge set of possibilities and the accuracy/aggressiveness tradeoff)