This is what I remember my father hammering into me from his Multician days when I was a child learning C and Lisp:

As one of the “triumvirate” running Multics with Fernando Corbató and Charlie Clingen, Neumann imposed the discipline for which Multics became known: No one was permitted to write a line of code until they had submitted a complete written English-language specification of the module’s behavior. Multics, Neumann later observed, solved the Y2K problem in 1965 and made stack buffer overflows impossible by construction.

From the ACM obituary.

I still have pages of specifications from my teens at home in a folder. Sigh, memories.