Creating economic value through code is currently an extremely complex process. When you consider what goes on between the ears of a basketball player, it is not trivial, but it can be transferred from one player to the next in a straightforward way.
Engineers must think beyond what is able to be communicated. Complex mental models of code and user experience that would take the best engineers longer to communicate than to instantiate.
At the top, you have a board of 10 people. Each one of these is making decisions and receiving orders for 100 people. This sort of structure is going to have some serious communication latency and a loss of data as it traverses the company heirachy.
That is quite normal :-)
And that's a strong argument to make. I might rather have five exceptional developers than 1000 average ones, taking into account communication latency and overhead, but if they're five surly, haughty, egotistical asshats who are engaged in a never-ending pissing contest, then no thanks.
The article totally misses a critical point: communications and coordination, which goes up exponentially with the number of players.
Group Intercommunication Formula: n(n − 1) / 2
1000 developers give 1000 * (1000 – 1) / 2 = 499500 channels of communication.
[edit] All Bill has established with his sports analogy is a team of 5-6 players that communicate effectively is better than a team of 5-6 players that don't communicate. That is a total strawman, it doesn't map to 1000 (software) players. [/edit]
In hockey, there are six or fewer players on the ice at one time. Would 1000 hockey players all on the ice at one time be better than six? Not a chance! They would not be able to move the puck because they would be tripping on each other. Even if the ice rink were expanded to hold them all, they still would suck as a team because their communications and coordination would be totally broken.
The same problem happens with software developers: 1000 developers working on one problem will trip over each other too. The only way 1000 developers will succeed at all is to decompose one big problem into 100-200 smaller problems that teams of 5-10 (funny how that works out) can then solve.