A non-empirical exchange functional based on an interpolation between two limits of electron density, slowly varying limit and asymptotic limit, is proposed. In the slowly varying limit, we follow the study by Kleinman from 1984 which considered the response of a free-electron gas to an external periodic potential, but further assume that the perturbing potential also induces Bragg diffraction of the Fermi electrons. The interpolation function is motivated by the exact exchange functional of a hydrogen atom. Combined with our recently proposed correlation functional, tests on 56 small molecules show that, for the first-row molecules, the exchange-correlation combo predicts the total energies four times more accurately than the presently available Quantum Monte Carlo results. For the second-row molecules, errors of the core electrons exchange energies can be corrected, leading to the most accurate first- and second-row molecular total energy predictions reported to date despite minimal computational efforts. The calculated bond energies, zero point energies, and dipole moments are also presented, which do not outperform other methods.