@danluu Fascinating read, but there are a fair few dead links in there now (partition testing, IBM’s SixthSense). I’m working on a PBT library that combines random sampling with pairwise/tway covering arrays of boundary/«interesting» values and optionally partition testing through specified «directions» (predicates that must be hit and are AND’ed with the property when reducing, each of which get a biased generator). It’s an incredibly interesting space, and I think a large part of the reason random testing hasn’t been adopted more widely is due to a combination of ergonomics and performance. Hypothesis is a great example of getting a lot of the ergonomics right.