Whose Now Is It?

 

Copyright: Sanjay Basu

I run into a timestamp problem most weeks, and it has nothing to do with philosophy.

A training run kicks off across a few hundred GPUs. The nodes agree, more or less, on what time it is, because we keep them disciplined with PTP and a grandmaster clock, and even then they drift by tens of nanoseconds against each other. A gradient all-reduce that finishes “at the same moment” on eight nodes did not finish at the same moment on any of them. The word “simultaneous” is a convenience we grant ourselves so the log files line up. When something goes wrong at that scale, the first thing an engineer learns is to stop trusting the wall clock and start trusting the causal order. What depended on what. Which message could not have arrived before which other message left.

I bring this up because I read a lovely piece this morning, from the Starts With a Bang newsletter, asking whether time really exists. It is a good question asked well, and the answer it lands on is that time is real because it is measurable, quantifiable, observable, and not pathological. You can put a number on it. Other numbers depend on it in ways you cannot pry apart. By the standards physics uses to decide what counts as real, time passes the test. I agree with all of that. It is also, I think, the least interesting thing you can say about time, and it quietly steps around the part that actually keeps people up at night.

Because nobody lies awake wondering whether their wristwatch is measuring something. They lie awake wondering why the watch only ever runs one way, and why this particular tick, the one happening now, feels different from all the ones filed away behind it.

The two questions hiding inside one word

When we say “time,” we are usually pointing at two different things and hoping context sorts them out.

The first is the coordinate. This is the t in your equations. It is the thing relativity mixes with space, the thing that dilates when you move fast and sags near a mass. This t is real in exactly the sense the article defends. It is also completely time-symmetric. Newton, Maxwell, Einstein’s field equations, the strong force, all of them run just as happily backward as forward. Film a planet orbiting and play it in reverse and you have not violated a single law. The coordinate does not know which way is later.

The second is the experience. The arrow. The overwhelming, non-negotiable sense that the past is fixed and the future is open, that eggs scramble but never unscramble, that I remember yesterday and not tomorrow. This is the time you feel. And here is the awkward fact the physics community has been polite about for a century. We do not have this second thing in our fundamental equations at all. We import it. We put entropy in by hand, through a low-entropy boundary condition at the beginning of the universe, and we let the second law do the rest.

The article is careful and honest about this. It says plainly that we do not know what causes the perceived arrow of time, and that the author’s hunch is the perceived arrow is not the same as the thermodynamic arrow. I want to sit in that gap, because I think it is the whole game, and I think there is a reason we keep failing to close it.

The move nobody wants to make

Here is the standard story. The fundamental laws are reversible. The universe started in a very low-entropy state. Entropy has climbed ever since. The direction of that climb is what we call the future. The arrow of time is the arrow of entropy, and the reason you remember the past and not the future is that memory is a physical record, and records are low-entropy structures that can only be laid down in the direction entropy increases.

It is a beautiful story and I have told it myself, on stage, with the ice-cube-in-the-drink example and everything. The ice never reassembles. Time only flows one way. The room applauds.

The trouble is that the story explains the wrong thing. It explains why systems tend toward equilibrium. It does not explain why I am sitting here, on the leading edge of a single moment, watching one outcome become fixed while others stop being available. Entropy is a statement about ensembles, about counting the number of microscopic arrangements that look the same from far away. It is a statement about my ignorance of the details. It says nothing about why there is a now, or why the now moves, or why it is this now and not some other one that I happen to be inside of.

Lee Smolin, quoted in the same article, put his finger on it better than most. He said the present moment, the thing everyone complains is missing from physics, is the transition from indefinite to definite. The moment when something that was quantum and undecided becomes classical and settled. That, he said, is what “now” is.

I think Smolin is exactly right about where to look, and I want to push on the word he leaves standing there unexamined. Definite for whom.

The QBist turn

I have written before that I am a QBist, which usually earns me a raised eyebrow and a question about whether I actually believe the moon is not there when nobody looks. So let me be plain about what the position is, because it changes this conversation completely.

QBism, quantum Bayesianism, takes the quantum state seriously as what it looks like it is. A catalog of one agent’s expectations about what they will experience if they act on the world. The wavefunction is not a thing floating in space waiting to be disturbed. It is a bookkeeping device that a particular observer uses to place bets on their next measurement. When the measurement happens and an outcome lands, the observer updates the catalog. That update, the thing the textbooks dress up as “collapse” and surround with paradox, is nothing more mysterious than a gambler revising the odds after the card is turned.

Notice what this does to Smolin’s transition from indefinite to definite. It stops being a cosmic event and becomes a personal one. There is no universe-wide switch that flips a superposition into a fact for everyone at once. There is an agent, taking an action, and experiencing an outcome, and the indefinite becomes definite for that agent, in that act. The “now” is not a property of spacetime. It is the moment an experience is created for someone who is in the world and acting on it.

This sounds like a retreat into solipsism and it is not, and the distinction matters. QBism is not saying reality is in your head. It is saying that the quantum formalism was never a description of reality from outside. It was always a first-person tool, a user’s manual for an agent embedded in a world that pushes back. The world is completely real. It is so real that it can surprise you, which is the entire point. You place your bet using the Born rule, the world answers, and sometimes the answer is not what you expected. That capacity of the world to answer back, to deliver an outcome you did not control, is as good a definition of “real and external” as I know.

So when the Starts With a Bang piece asks whether time exists and answers yes because you can measure it, a QBist nods and then asks the follow-up the measurement framing invites. Who is doing the measuring. Because measurement is not a thing the universe does to itself. It is a thing an agent does. And the moment of measurement, the click of the detector, the settling of the indefinite into the definite, is the only place in all of physics where the now shows up naturally, without being smuggled in through a boundary condition.

Why this dissolves the argument instead of winning it

I am not claiming QBism proves the arrow of time. I want to flag that clearly, because it would be easy to read this as a triumphant solution and it is not one.

What the QBist move does is smaller and, to me, more useful. It relocates the problem. It says the reason we cannot find the arrow of time in the fundamental laws is that we have been looking in the wrong ledger. The laws describe the structure of the world that any agent will find waiting for them. They are third-person, view-from-nowhere, and view-from-nowhere is exactly where the now cannot live, because a now is always somebody’s now. Asking the block universe where the present moment is located is like asking a map where “here” is. The map does not have a “here.” “Here” is a fact about the person holding it.

On this reading, the block universe and the flowing present are not competing theories that one of them has to lose. They are two different documents. One is the third-person structure, timeless and symmetric, the thing the equations describe. The other is the first-person experience, oriented and flowing, the thing an embedded agent lives through as they turn their expectations into outcomes one action at a time. The mistake of the last hundred years, and I include a younger version of myself in this, is to demand that the first-person now be derivable from the third-person structure. It is not derivable from it because it was never in it. It is what you get when an agent enters the structure and starts acting.

I hold this loosely. There are real objections. If every agent has their own now, how do our nows agree well enough to schedule a meeting, or run a distributed training job across nodes that must synchronize. QBists have answers, and they lean on the fact that agents share a world and can compare notes afterward, which is precisely the relativity-of-simultaneity story dressed in different clothes. But “we compare notes afterward” is doing a lot of work in that sentence, and I do not think anyone has fully cashed it out. This is one of those places where honest uncertainty is the correct final position and pretending otherwise is how you fool yourself.

Back to the machines

I promised myself I would not write a pure philosophy piece, because the datacenter keeps me honest and the datacenter has opinions about time.

Every distributed system I have ever operated has quietly rediscovered relativity. There is no global “now” in a cluster. There is only the causal order, the record of which event could have influenced which other event, reconstructed after the fact from timestamps we already know are lies at the nanosecond level. Leslie Lamport built most of distributed computing on exactly this insight in 1978, that in a system with no shared clock, the only time that matters is the partial order of who could have heard from whom. Physicists call it the light cone. Distributed systems engineers call it happens-before. It is the same idea, and both fields arrived at it because both were forced to give up on a universal present.

The agents I build have the same problem in a sharper form. An AI agent maintaining memory across a long task is doing something that looks eerily like laying down a thermodynamic arrow. It reads its context, it acts, it writes a record, and that record is what lets a later version of itself distinguish past from future. Take the memory away and the agent has no arrow at all. Every moment is the first moment. It is stateless, which is another way of saying it is timeless, which is another way of saying it does not have a now. The arrow of time, for a machine, is a design decision. You get it by choosing to persist state, and you pay for it in exactly the currency the second law demands, in energy spent and heat dumped to make an irreversible record. Landauer told us in 1961 that erasing a bit has a minimum thermodynamic cost, and every agent that remembers is buying its sense of time on that account.

I find this clarifying rather than reductive. It suggests the arrow of time, at least the version we care about, the version where the past is fixed and the future is open, might be less a fundamental feature of the universe and more a feature of what it is like to be a thing that records its own history while embedded in a world it cannot fully predict. Which is to say it might be a feature of being an agent. Human, or otherwise.

Where I actually land

Does time exist. Yes, in the boring sense, and the Starts With a Bang piece defends that sense correctly. The coordinate is real, the dilation is real, the entropy is real, and anyone telling you time is a total illusion is selling something. Most probably weed.

But the question people are really asking, the one about the now and the arrow and why this moment feels like the moving front of a fire, is not a question about whether time is measurable. It is a question about what it is like to be a measuring thing. And that question does not get answered by pointing at a clock, because the clock is part of the third-person furniture, and the now was never furniture. The now is the act of turning the indefinite into the definite, and that act belongs to an agent, not to the cosmos.

I could be wrong about all of this. QBism might turn out to be an elegant dead end, a way of feeling better about the measurement problem without solving it. The thermodynamic arrow and the experiential arrow might turn out to be the same thing after all, and I will have written several hundred words insisting on a distinction that evaporates. I have changed my mind on smaller things and larger.

Here is the question I cannot shake, and I will leave it with you rather than pretend to answer it. If the now is always somebody’s now, and the arrow of time is the price an agent pays to remember, then what happens to time in a universe with no agents in it at all. Does it flow. Or does it just sit there, symmetric and complete and waiting, like a map with no one holding it, for the first thing that can say “here” and mean it.