Until now, computationally designed enzymes exhibited low catalytic rates1-5 and required intensive experimental optimization to reach activity levels observed in comparable natural enzymes5-9. These results exposed limitations in design methodology and suggested critical gaps in our understanding of the fundamentals of biocatalysis10,11. We present a fully computational workflow for designing efficient enzymes in TIM-barrel folds using backbone fragments from natural proteins and without requiring optimization by mutant-library screening. Three Kemp eliminase designs exhibit efficiencies greater than 2,000 M-1 s-1. The most efficient shows more than 140 mutations from any natural protein, including a novel active site. It exhibits high stability (greater than 85 °C) and remarkable catalytic efficiency (12,700 M-1 s-1) and rate (2.8 s-1), surpassing previous computational designs by two orders of magnitude1-5. Furthermore, designing a residue considered essential in all previous Kemp eliminase designs increases efficiency to more than 105 M-1 s-1 and rate to 30 s-1, achieving catalytic parameters comparable to natural enzymes and challenging fundamental biocatalytic assumptions. By overcoming limitations in design methodology11, our strategy enables programming stable, high-efficiency, new-to-nature enzymes through a minimal experimental effort.