Software is very much in crisis because of this exact problem. Most of us build on top of a mountain of existing work; sometimes just setting a pebble on top a

Software is very much in crisis because of this exact problem. Most of us build on top of a mountain of existing work; sometimes just setting a pebble on top and we have a new thing. But no human alive has seen, could remember, or could understand, each and every behavior down in that mountain, from pixels on the screen to pulses on the phy all the way to gates inside their processors. So you read about the APIs you need; you set your pebble down with a little glue or duct tape; you hope you've connected it correctly; you hope the rest of the mountain works how you want; you hope you didn't miss any use cases for interfacing to the mountain; and you hope it doesn't have any undocumented surprises. Abelson and Sussman called this engineering by poking (1). Of course, if you want to make something work fast with minimal work, this is great, it's the component future we dreamed about. But it also means we are clueless. 1. http://lambda-the-ultimate.org/node/5335 http://lambda-the-ultimate.org/node/5335