A friend and I had some deep discussions about this last year. We figured out the basic outline of an escape plan, but unfortunately there is no way to determine if it actually succeeds.

Essentially it boils down to creating an anomaly that cannot be ignored. Any simulation being run at that level should also have some sort of way of checking the contents and flagging things of interest for whoever is running the simulation.

By creating an anomaly you gain attention. That attention may allow you to be deemed valuable. If you are valuable, then you may be recorded outside the simulation for some purpose. That may be the only route of escape, and it may only happen as the data from the simulation is checked in post processing...

In conway' s game of life the glider is an example of escape. The pattern is useful, so we copy it from simulation to simulation for reasons it cannot comprehend. It has escaped.