I will ask the usual question when a new ML implementation appears on my radar: is it `standard' enough that it could be used to just check for correctness of code during development?