Despite promising advancements in the field of computational protein design, computational methods are not able to reliably identify viable therapies for the treatment of human disease. This is due to the fact that the prevailing paradigm has relied on the development of compounds with strong binding affinities to molecular targets, a task that is not readily achievable by computational methods. To make progress in this pursuit, we make the following conceptual leap: novel drug mechanisms of action need to be devised according to the capabilities of current computational design technology. This represents a paradigm shift from the prevailing narrative of computational methods as auxiliary tools, to one where computational and experimental approaches evolve together according to the state-of-the-art of each domain. By bridging these two technological arenas, the long-sought after capability of push-button computational drug design can be realized.