Take some finite program and an infinite amount of memory. Now run that program, and you can build increasingly complex spaces in that memory. This is basically how automata work. If the finite program is the code of the universe, "now" depends on executing the program up until now, so it's of little use in giving us a shortcut to understand now (but we'd still like to know what this program is!). Also, just knowing a snapshot of now doesn't give us the future without knowing the program (and in any case, the time it takes to compute the future is at least the amount of time it takes to reach that future).
Who says the computation has to run inside of time?
Consider our spacetime as a four-dimensional space, the pixels of which are set by a computation running in a different dimension of "time." The number of steps required to generate the final output is arbitrary and unrelated to anything we observe, and there's no contradiction in supposing that, say, information in our future could be an input to an iterated algorithm that determines our present.
