I think what they may've been referring to is something like the Curry-Howard correspondence, which describes the mapping between computer programs and mathematical proofs. Mathematical proofs can't be patented, and it could be taken to imply that (at least some) programs/algorithms shouldn't be eligible for patent protection.