> You will have bad data in prod. Users will do unexpected things. You won’t always be able to reach S3. And, this is all true even assuming the software is wri

> You will have bad data in prod. Users will do unexpected things. You won’t always be able to reach S3. And, this is all true even assuming the software is written perfectly to spec. Surprisingly this nonsense happens in a discrete setting where we can (in theory, of course) enumerate all possible inputs. Again surprisingly we cannot find effective way to automagically partition input space into equivalence classes to assist us in testing. Novel works in conjunction with our brains and it is not surprising that there are no definitive rules or “good novel equation”. But programs implement defined algorithms and super predictable. Yet creating them feels like writing a novel. I agree with your statement but at the same time it feels wrong that we find ourselves in such position.