We consider the Online Boolean Matrix-Vector Multiplication (OMV) problem studied by Henzinger et al. [STOC'15]: given an nfin Boolean matrix M, we receive n Boolean vectors v1; : : : ;vn one at a time, and are required to output Mvi (over the Boolean semiring) before seeing the vector vi+1, for all i. Previous known algorithms for this problem are combinatorial, running in O(n ³=log ² n) time. Henzinger et al. conjecture there is no O(n ³-ϵ ) time algorithm for OMV, for all e > 0; their OMV conjecture is shown to imply strong hardness results for many basic dynamic problems. We give a substantially faster method for computing OMV, running in n ³=2W( p logn) randomized time. In fact, after seeing 2w( p logn) vectors, we already achieve n2=2W( p logn) amortized time for matrix-vector multiplication. Our approach gives a way to reduce matrix-vector multiplication to solving a version of the Orthogonal Vectors problem, which in turn reduces to "small" algebraic matrix-matrix multiplication. Applications include faster independent set detection, partial match retrieval, and 2-CNF evaluation. We also show how a modification of our method gives a cell probe data structure for OMV with worst case O(n7=4= p w) time per query vector, where w is the word size. This result rules out an unconditional proof of the OMV conjecture using purely information-theoretic arguments.