Consider the Prisoner's Dilemma game played for one round. The only
Nash Equilibrium is for both players to defect even though they would
both gain if they both cooperated. Tennenholtz (GEB 2004) consider
program equilibria where each player provides a program that can look
at the other player's program. Tennenholtz shows that cooperation is
now possible in this model.

Tennenholtz and others use a limited computation model that, for
example, lacks universal simulation. We look at a new model that
allows for arbitrary Turing machines but discounts the payoffs based
on running time. We recover the results to Tennenholtz and others in
this model and show our model also handles some games that have no
program equilibria in the old sense. Finally we show how discounted
time may allow us to understand strategies used in computationally
difficult games like chess.