The Arrow of Time — equal before the laws, not the record — rendered package. Rendered from entropy-arrow.md; explanatory companion, changes zero gate statuses.

The Arrow of Time — equal before the laws, not the record

Everything on these pages is conditional on the same small set of roots and measured anchors, and the arrow of time is no exception. This page is not new physics. It closes no open problem and changes no gate status. It takes the most familiar asymmetry in existence — the past is fixed, the future is not; eggs break and do not unbreak — and sorts it honestly into the same ledger everything else on these pages lives in. The result is not an answer to "why does time flow?" It is a re-typing of the question, and the re-typing is the whole content.

$$\boxed{\text{The laws do not point. The record does. The arrow is a boundary fact, not a law.}}$$

The "second law of thermodynamics" is usually spoken as one thing. It is three things, of three entirely different kinds, and blurring them is where most of the confusion about time comes from. One is derivable. One is an emergent lemma. And one — the arrow itself — is neither: it is a contingent, measured fact about this universe, and treating it as anything grander is exactly the kind of over-claim these pages exist to refuse.

Carried from the sibling pages, and binding here too: dissolved $\neq$ solved — see Dissolutions.

Three claims that must not be blurred

Claim What kind of thing it is What backs it
Entropy is a well-defined count. Derivable in principle — conditional on the Granularity posit. There is a finite operational floor under any distinction (Granularity). A finite floor means a finite region at finite energy holds only finitely many physically distinguishable states. Where states can be counted, entropy is a count — not a metaphor, not a mood.
Entropy (overwhelmingly) does not decrease. An emergent lemma — a theorem-shaped consequence, not a posited law. Given the count, the dynamics, and coarse-grained records, non-decrease follows by typicality: among all the ways a system can be arranged, the overwhelming majority of what happens next looks like "entropy stays or grows." Nobody has to legislate it.
Time has an arrow. Not a law. Not derivable. A contingent boundary condition — a measured fact about this universe. The observed low-entropy past: the early universe was extraordinarily smooth and simple, and we have the record — the cosmic microwave background is its surviving photograph.

The first row costs nothing new: it rides on structure already granted. The second row is standard physics, cited here rather than re-derived — an honest IOU in this system's own books (see Honest status below). The third row is where every popular account goes wrong, and it is the reason this page exists.

Why the arrow cannot be derived — the lemma is symmetric, the record is not

Here is the trap, stated plainly enough to walk around it.

The typicality lemma — "from any ordinary state, entropy almost certainly increases" — is time-symmetric. The microscopic laws it runs on treat both directions of time on essentially equal footing. So run the very same argument toward the past, and it "predicts" with the same confidence that entropy was higher yesterday: that your memories assembled from noise, that the half-melted ice in your glass condensed out of lukewarm water a moment ago.

That prediction is false. Spectacularly, checkably false — every record you own says entropy was lower in the past, all the way back to the smooth early universe.

So the lemma, by itself, cannot be the arrow. Something breaks the tie between the two time directions, and it is not the laws — the laws are even-handed. What breaks the tie is a boundary condition: the universe started (in the record-ordered sense) in a state of extraordinarily low entropy. Given that one contingent fact, everything follows — entropy climbs away from the boundary, records accumulate away from the boundary, and the direction of climb is what we call "the future." Remove that fact, and the laws alone hand you no arrow at all.

This is the paradigm case of a distinction drawn on the sibling page One, But Separate:

$$\boxed{\text{Both directions of time are equal before the laws. Only one has the record.}}$$

Relativity-style even-handedness holds at the level of the laws: no direction is preferred by them. The asymmetry lives entirely in the record — in which direction the finite, physical records actually exist. The arrow is not written into the rulebook. It is written into the boundary, and read off the records.

(One honest footnote: the known microscopic laws are not perfectly time-symmetric — the weak interaction violates T by a tiny, precisely measured amount. That violation is real and precisely measured, but it cannot source the thermodynamic arrow: it is reversible in the deeper CPT sense and has nothing like the size or structure needed to make eggs unbreak in only one direction. The tie-breaker is still the boundary.)

What paying for the arrow honestly looks like

On these pages, a fact that cannot be derived and is observed has exactly one honest disposition: it is a measured input — named, recorded, and counted, like every other measured anchor in the system.

The low-entropy past qualifies on every test:

And the honest current status, stated plainly: as of this page, nothing in this system's accounting actually consumes the low-entropy past. It is observed, named, and — here — used by nothing. It therefore costs nothing yet. The moment any account on these pages leans on it, it gets counted as one more measured input, in the open, like every other. No silent consumption.

The honest guard — observer arguments derive only the perceived arrow

There is a family of seductive arguments that must be flagged, because they sound like derivations of the arrow and are not.

They go like this: any observer is a record-keeping physical object; record-keeping is itself an entropy-increasing process; therefore any observer will find records of the "past" and none of the "future," and will perceive time as flowing — whatever the universe is doing. Some versions add: so the arrow needs no explanation at all.

The guard: these arguments derive the perceived arrow, not the arrow. They show — correctly, and it is worth knowing — that no observer could ever experience time as arrowless. What they do not show is why this universe actually has a low-entropy boundary for observers to exist downstream of. An argument that explains why any witness would report $X$ has not explained why $X$ is there to be witnessed. Selecting the answer out of the observer's situation is not the same as deriving it from the laws — and every observer-based "derivation" of the arrow, unpacked carefully, assumes the low-entropy boundary at some step or delivers only the perception.

So the observer arguments are kept, honestly labelled: they are a true account of experience, and zero progress on the boundary condition itself.

The question that re-types

"Why do the laws of physics prefer one direction of time?"

They don't. The question presupposes an asymmetry in the laws, and no such asymmetry (of the required kind) exists. Asked of the laws, the question is aimed at the wrong object — and in that form it does not get answered; it dissolves, conditionally, the way malformed demands do on these pages. What remains after the re-typing is a different question, well-posed and genuinely open:

"Why was the boundary low-entropy?"

That question is real, and this page does not answer it. Today it has the same standing as any deep contingent fact: the honest endpoint currently in view is that the low-entropy past is a measured feature of this universe — recorded, not explained. Whether something deeper will one day derive it is left fully open. Dissolved $\neq$ solved: re-typing the first question tells you about the question. It settles nothing about the second.

The falsifier

A page that only re-types questions still has to be breakable. So, published plainly:

If the low-entropy past is ever derived from the laws plus the roots alone — no boundary input assumed anywhere in the argument, checked line by line — this page's central claim fails. The arrow would move from measured fact to derived result, the input would be struck from the ledger, and this page would be rewritten, not defended. (That would be a better outcome than this page. Nothing here hopes it stays true.)

And a second, structural one, aimed at the first row of the table:

If a finite region at finite energy is shown to support infinitely many physically distinguishable states — records finer than any floor — the counting claim fails, and with it this page's ground for treating entropy as a well-defined count rather than a convention.

A third, for the lemma: the typicality reading predicts non-decrease as overwhelming, not exceptionless — tiny, fleeting reversals are expected and observed. A reproducible, macroscopic, sustained entropy decrease in an isolated system would break the lemma as stated, and no appeal to rarity would be allowed to absorb it.

Honest status

Anti-claims (held on this page):

Cross-links