I was thinking about Simon Newcomb, the eminent astronomer/mathematician who mathematically "proved" that human mechanical flight was forever impractical a few months before the Wright brothers went ahead and did it.
(Look into what he might have missed, I think he was leaning on the square-cube law: as a machine grows, its volume and weight increase cubically, while its wing area increases only quadratically, by the square.)
I had a similar incident in real life, where my friend EB measured an overstuffed chair and pronounced it impossible to get through a small door and into a back room, and then a few days later my friend JZ and I just did it. (Here, EB really didn't account for the "pivot" of it all - must have missed that episode of Friends.)
That's one of the dangers of reductionism; a model captures what it can, and ignores the rest.
I see a parallel in software testing... unit tests are fast, focused, cheap to run and now incredibly tempting to just throw an AI at to produce. But the success or failure of a working system is an emergent property of all those units working together in a context provided by all the other units... and browsers, networks, external systems, and of course users. Integration and end to end testing are expensive but more important than ever.
So test the units, but don't forget to test the pivot.
No comments:
Post a Comment