Electrocatalysts support crucial industrial processes and emerging decarbonization technologies, but their design is hindered by structural and compositional changes during operation, especially at application-relevant current densities. Here we use operando X-ray spectroscopy and modelling to track, and eventually direct, the reconstruction of iron sulfides and oxides for the oxygen evolution reaction. We show that inappropriate activation protocols lead to uncontrollable Fe oxidation and irreversible catalyst degradation, compromising stability and reliability and precluding predictive design. Based on these, we develop activation programming strategies that, considering the thermodynamics and kinetics of surface reconstruction, offer control over precatalyst oxidation. This enables reliable predictions and the design of active and stable electrocatalysts. In a NixFe1−xS2 model system, this leads to a threefold improvement in durability after programmed activation, with a cell degradation rate of 0.12 mV h−1 over 550 h (standard operation: 0.29 mV h−1, constrained to 200 h), in an anion exchange membrane water electrolyser operating at 1 A cm−2. This work bridges predictive modelling and experimental design, improving the electrocatalyst reliability for industrial water electrolysis and beyond at high current densities.