A few months before launching the Thinking Machines company, Hillis discussed his ideas of building a massively parallel machine with Richard Feynman over lunch. Feynman pronounced the project ``dopey'' but then promptly volunteered to work for the company the following summer. Demanding ``something real to do'' when he arrived, he helped analyze the interconnection network of the first machine, determining how much short term memory each switch in the network would need....
His use of continuous mathematics as an analytic tool struck Hillis and the other computer designers as eccentric when they first heard it, but Feynman's analysis proved to be correct.