[whatshisface]

That's essentially half my point. The organization might be passable at "building rockets that are moon-capable," but it's going to be worse at each individual specialized task than the individual experts that make it up. One of the C programmers is likely to look at the architecture of the control system and think "that's a dumb way to architect a control system," and he'll probably be right. The same will be true for every subsystem, and even the integrated assembly. However unless someone figures out how to reduce communication costs, that's the best anyone can do; the alternative would be to have a great control system with no rocket.

[yikesshescute]

Is the implication that a company can spend more time and money to get each subsystem to subsystem-expert-acceptance level, or is there something about current organizational practice that makes even this impossible?

[whatshisface]

Imagine two experts advising a leader. They're not going to fully communicate all of their expertise to him, so his decisions on topic A will probably be worse than expert A, and likewise for topic/expert B. He's going to make better decisions on joint A+B tasks that require combined and balanced knowledge, but both experts will (not incorrectly) see him as "not knowing as much about A/B as I do."

You can substitute the individual leader for a group whose consensus must be obtained and get another parable illustrating the same principle.

[yikesshescute]

So, the leader is an expert on “joint A+B tasks”? Then provided he, the leader, can identify, A tasks, B tasks and A+B tasks and correctly defer on the first two sets, won’t we get optimal decision making?

And are we sure that there really are A+B tasks?