Traditional mechanism design engineers a game G so that a desired property P holds at one or more of G's equilibria. But due to equilibrium selection and collusion, as well as computational complexity and privacy issues, it may fail to guarantee its desired properties.

We thus put forward an alternative, equilibrium-less, and more robust approach to mechanism design, and exemplify it for several classical problems.

(Based on various recent works with Jing Chen, Avinatan Hassidim, and Paul Valiant)