I agree with basically all of your article. It certainly explains the dichotomy of physical laws allowing a spacetime to be "played backwards", while our experience must always be that of it being "played forwards"-- the time reversal of learning something is forgetting it, and this would invert the topological sort of knowledge dependency that places slices of our experience in the "future" or the "past". As such an observer in an isolated part of the universe which was being "played backwards" would label our future as their past.

If we buy into the Second Law, it makes sense to me that there would be some direction where there are more messy entanglements between all the parts of the universe, and in the other direction there would be fewer, and we would name the first direction "the future" and the second direction "the past", since there would be information embedded in all those mutual entanglements that would over-constrain states in the past and under-constrain states in the future.

What I don't get is what prevents me from setting up a circumstance where a measurement I make in the present is correlated with the result of (say) a coin flip in the future. That is easy to do if the coin flip is in the past, but impossible the other way around. I agree that by your logic if I were getting more information on the whole from the future, then I would almost definition label that "the past". But why not even one bit from the future? I can make a meaningful mark on a particle at time t0 and read it at t1, but not if t0 > t1. How come?

> there would be some direction where there are more messy entanglements between all the parts of the universe

Yes, that's exactly right, except for the word "messy". It's not "more messy", it's just "more".

> what prevents me from setting up a circumstance where a measurement I make in the present is correlated with the result of (say) a coin flip in the future

Nothing. If you flip a coin with enough precision you can make it come up reliably on one side or the other. ("But that's cheating!" you say. "I want the flip to be random." Well, you can't have it both ways: if the flip is random, then by definition it's not going to be correlated with anything in the past!)

> why not even one bit from the future

Because then it wouldn't be the future.

Think about this: how would you distinguish "receiving a bit from the future", which is apparently not possible, and "making a reliable prediction about the future", which is possible in many cases?

> I can make a meaningful mark on a particle at time t0 and read it at t1, but not if t0 > t1. How come?

You can't actually "make a mark" on a particle the way you can on a classical object. You can prepare a particle in a particular quantum state, but that's not the same as putting a mark on it.

Yes, of course. That was loose wording for "affect a particle a way that is meaningful", i.e. prepare it. :)

> "But that's cheating!" you say. "I want the flip to be random."

Actually, no, I'd rather the bit be useful!

> If you flip a coin with enough precision you can make it come up reliably on one side or the other.

Sure, if I know enough about the universe at some moment in time, by unitarity, I know its state at all other times. Having enough information to run the laws of physics forward to compute the outcome of the 2020 super bowl coin flip isn't much different, then, than running them backward to get the 2019 flip (and just as impractical). But what is interesting is that I could correlate (say) the spin of an electron with the result of the flip and read it later to tell what the flip was, but not earlier to tell what the flip will be.

(Obviously information about the past coin flip is available in many more places than just in my prepared particle, but I don't need anything more than that one qubit to precisely know the result. That's a big difference from needing to know everything in the coin's past light cone! )

If we peel away the human-imposed notions of time and causality (as you do in your article), and see spacetime as a "block" with microscopic time symmetry, or perhaps even further dismantled into only basis states of Hilbert space, it's still obscure to me why particles-- even individual ones-- seem to "carry" information only from the past and not from the future.

On an intuitive level it seems perfectly natural. On the level of (time-symmetrical) fundamental physics, I can't pin down why it would be.

Yes, but these details matter. Preparing a particle in a quantum state is fundamentally different from making an identifying mark on a classical object.

> > If you flip a coin with enough precision you can make it come up reliably on one side or the other.

> Sure, if I know enough about the universe at some moment in time, by unitarity, I know its state at all other times.

That's not what I meant. I'm not talking about trying to measure the state of the coin and the flipping apparatus in order to predict the outcome, I'm talking about building a precision flipping apparatus that allows you to control the outcome.

> I could correlate (say) the spin of an electron with the result of the flip and read it later to tell what the flip was, but not earlier to tell what the flip will be.

Actually, you can do both, and the procedures are essentially identical: to do the former, you look at the coin and manipulate the electron state to match. To do the latter you look at the electron and manipulate the state of the coin to match. Easy-peasy.

Is that not what you wanted? If not, why? (Remember that when I suggested you would want the coin flip to be random, you denied it.)