I think Kent was being sarcastic - his point is that it's rather pointless to optimize until the system is producing correct results. He's also famous for the "Make it work, make it right, make it fast" quote.
A quick scholar.google.com search found a match in Byte magazine from 1983:
> Furthermore, many C environments contain
measurement tools that enable the programmer to identify these critical sections easily. But the strategy is definitely: first make it work, then make it right, and, finally, make it fast.
