Sometimes (quite often actually) there are functions that aren't trivial to implement but at the same time it is almost trivial to come up with concrete input/output pairs for them.
GoodGuesser lets you derive a function based on the examples you provide.
Yes that's exactly right. My library isn't doing anything super smart (just statistics 101) but if you have to do some awkward heuristic in your code (to "eyeball" some approximate parameters) it's better to just supply some input/output examples as part of your code (which GoodGuesser autogenerates) then just have a computer approximate the ideal parameters using multiple linear regressions. That will be a lot easier to maintain than some sort of hacky heuristic.
Plus, once you've refactored your logic as input/output pairs, It'll be easier in the future to do fancier stuff with your code if you want (deep neural nets, etc)
Sometimes (quite often actually) there are functions that aren't trivial to implement but at the same time it is almost trivial to come up with concrete input/output pairs for them.
GoodGuesser lets you derive a function based on the examples you provide.
Related twitter thread from february: https://twitter.com/lisperati/status/1492165176451387401