Not GP, but I interpreted it as an argument to innovate/proliferate implementations early and often, but to standardize minimally and as late as possible.