He's probably thinking of Paul Feyerabend; the best example for Feyerabend is probably his studies of Galileo - demonstrating that Galileo's observations did not prove his theories, his theories made poorer predictions than geocentrism, replication of his results often failed, and so on, and that Galileo succeeded more on account of non-empirical reasons such as theoretical elegance and social connections than on the then-merits of his theory.