A passive machine on a surface (of gravitationaly attracting sphere = ideal Earth) cannot be in a stable or meta stable state with only two supports.
A di-pod cannot span an area, in a way that it's center of gravity stays within if tilted.
unlike the doll below, which has always a projected area around it's point of support under it's center of gravity - a two-legged system spans only a line, which is easy to cross.
Self-balancing system should revert to it's state after a small (and to be defined, in the example below it would be $x_a - x_b$) dicsplacement. In terms of a potential this is the case for stable and meta-stable states.
In the picture the state of your system is related to potential (y-axis), and some abstract position is at the x-axis: e.g. Your system is staying "upright" at $x_a$, it will go back to $x_a$ if pushed with less effort than $\Delta U$, and will be "falling down" untill it reaches next stable position at $x_c$ "lying down"
A bag hook could be considered stable virtually having only one support, but it's not an option, for a) moving machine b) on a surface
Upon further reflection - the hook is actually hanging on a table, which has at least three supports itself.