<p>Electrocatalysis plays a central role in clean energy conversion and sustainable technologies. However, the trade-off between activity and stability of electrocatalysts largely hinders their practical applications, notably in the oxygen evolution reaction for producing hydrogen and solar fuels. Here we report a steam-assisted synthesis armed with machine learning screening of an integrated Ru/TiMnO_x electrode, featuring intrinsic metal-support interactions. These atomic-scale interactions with self-healing capabilities radically address the activity-stability dilemma across all pH levels. Consequently, the Ru/TiMnO_x electrode demonstrate enhanced mass activities—48.5×, 112.8×, and 74.6× higher than benchmark RuO₂ under acidic, neutral, and alkaline conditions, respectively. Notably, it achieves stable operation for up to 3,000 h, representing a multi-fold stability improvement comparable to other state-of-the-art catalysts. The breakthrough in activity-stability limitations highlights the potential of intrinsic metal-support interactions for enhancing electrocatalysis and heterogeneous catalysis in diverse applications.</p>