It doesn’t feel like news, yeah.

I would emphasize, though, that getting clearly defined input _that remains stable_ is hard. Often something is discovered during implementation of a task that informs changes in other task definitions. A parallel system has to deal with this or the results of the parallel tasks diverge.