Chapter 13: Astronomia — Cosmic-Scale Entropy

Why is the night sky dark?

The question sounds childish, yet it puzzled astronomers for centuries. If the universe is infinite and filled with stars, every line of sight should eventually hit one, the way every sightline in a dense forest eventually hits a tree trunk. The sky should blaze in every direction, day and night, forever. Instead: darkness punctuated by points of light.

Astronomers call this Olbers’ paradox, named for the German astronomer Heinrich Olbers, who restated it in 1823.¹⁸ Its resolution tells us something important about the universe’s thermal history.

The sky is dark because the universe is young.

Light takes time to travel. The most distant objects we can see are those whose light has had time to reach us since the universe began. Beyond that horizon, more stars may exist, yet their light has not arrived. Stars themselves have finite lifetimes, so even within the horizon there has not been time to fill every sightline with starlight. These two facts, the finite age of the cosmos and the finite lifetimes of stars, are the principal reasons the sky is dark. Expansion adds a secondary effect: the light that does arrive from the most distant sources is stretched toward longer wavelengths we cannot see, a phenomenon physicists call redshift.

The darkness is evidence: something happened. The cosmos has a beginning, an evolution, an arrow of time pointing in one direction.

That direction is entropy.

The previous chapter traced this arrow through molecular chirality and parity violation; here it reaches cosmic scale.

Figure 13.1: Static Universe: in an infinite, unchanging cosmos every line of sight eventually hits a star, making the night sky blazing bright. Young, Expanding Universe: a finite age and finite stellar lifetimes leave most sightlines empty, and expansion redshifts the most distant light beyond visibility, giving us the dark sky we actually see.

The Low-Entropy Beginning

A puzzle that should trouble you:

The Second Law says entropy increases. The universe began 13.8 billion years ago.¹⁹ If entropy has been increasing ever since, the universe must have started in a state of extraordinarily low entropy.

Correct, but counterintuitive.

Look at the cosmic microwave background radiation, usually called the CMB: a snapshot from when the cosmos was just 380,000 years old. A nearly uniform glow. Matter and energy spread almost perfectly evenly in all directions. No stars, no galaxies, no structure. Smooth, hot plasma filling all of space.

This looks like high entropy. Uniformity is the signature of thermal equilibrium: everything spread out, no gradients.

For ordinary thermodynamics, yes. The early universe was different, because gravity dominated, and gravity reverses the usual thermodynamic intuition.

Gravitational Entropy

Intuition fails here, so take this slowly. In an ordinary gas, molecules can be spread out in many ways and clumped together in few. Spreading wins. That is the familiar story of cream dissolving in coffee: it disperses because dispersed states vastly outnumber concentrated ones.

Add gravity, and the picture inverts. Imagine a crowd in a park: without attraction, people spread out evenly. Now give everyone a gentle pull toward anyone nearby. They clump, and the clumps grow.

Clumps pull more matter in, making clumped states increasingly probable. The statistics flip: for gravitating systems, clumping increases entropy.⁵³⁴ The more concentrated a region becomes, the more it pulls, the more concentrated it gets. Clumping is the path toward equilibrium, the way a ball rolls downhill toward rest.

A perfectly uniform distribution of gravitating matter is unstable, like a pencil balanced on its point. The slightest density fluctuation grows, pulling matter into clumps that pull in more matter. Gravity is how the universe builds structure from smoothness.

This is how galaxies, stars, and planets form. The structure we see, all of it, is gravitational entropy increasing.

The early universe, so smooth and uniform, was in a state of extraordinarily low gravitational entropy: a coiled spring, loaded with potential energy and ready to release. The 13.8 billion years since have been the gradual uncoiling: matter clumping, structure forming, gradients spending themselves. We are living in the middle of that release.

The Cosmic Microwave Background

The CMB is the oldest light in the universe. It was released when the cosmos cooled enough for electrons to bind to nuclei, making space transparent for the first time: an event astronomers call recombination.

Before this moment, the universe was a plasma: a soup of charged particles scattering light in every direction, making the cosmos opaque, like fog so thick you cannot see your hand.

When the plasma cooled and formed neutral atoms, the fog cleared and the photons escaped. They have been traveling ever since, stretched by the expansion of space from visible light to microwaves. Today they fill the universe with a faint glow at 2.73 kelvin, just under three degrees above absolute zero.²⁰

The CMB is almost perfectly uniform: the same temperature in every direction to within one part in 100,000. This uniformity signifies early equilibrium, a state where the universe was so hot and dense that everything was in thermal contact. That 2.73-kelvin glow is the actual temperature of the cosmos today, about minus 270 degrees Celsius.

Those tiny variations, one part in 100,000, are the seeds of everything. They began as quantum vacuum fluctuations: random jitters in the energy of empty space during the first instant after the Big Bang. These ripples were unimaginably small, smaller than an atom.

Then came inflation. In a trillionth of a trillionth of a trillionth of a second, the observable universe doubled in size roughly eighty times, stretching quantum-scale ripples to cosmic proportions.

Gravity did the rest. Denser regions pulled in more matter, became denser still, and collapsed into the first stars and galaxies.

The CMB is a snapshot of the low-entropy beginning. The structures we see today are what happened when entropy did its work.

That snapshot has a testable consequence. If the universe has been expanding and cooling since recombination, its temperature at any earlier epoch should scale with the expansion: T = T₀(1 + z), where z is the redshift and T₀ is today’s 2.73 kelvin. That z is a stretching factor read straight off the light: at z = 1 the waves arrive twice as long as they set out, at z = 2 three times as long, so a source at z = 6 is showing us light from a cosmos seven times more compact than today’s. A universe half its present size should have been twice as hot. The prediction is straightforward. Confirming it requires a thermometer that works across billions of light-years.

Nature provides one. Quasars, the intensely luminous cores of distant active galaxies, blast light across the observable universe. That light passes through foreground galaxies, where sparse molecules of hydrogen cyanide and carbon monoxide float in gas so thin that the only thing warming them is the CMB itself. The molecules equilibrate with the background radiation and imprint its temperature on the quasar light as characteristic absorption fingerprints. In 2013, Muller and colleagues measured a CMB temperature of 5.08 ± 0.10 kelvin at redshift 0.89, roughly seven billion years ago.⁵³⁵ In 2025, Kotani and colleagues repeated the measurement with ALMA, obtaining 5.13 ± 0.06 kelvin: 40% more precise and within the measurement uncertainty of the predicted 5.14 kelvin.⁵³⁶

A complementary technique reaches further back. Riechers and colleagues observed a massive starburst galaxy at redshift 6.34, just 880 million years after the Big Bang, where a cloud of cold water vapor cast a measurable shadow on the CMB behind it. The degree of darkening revealed a temperature between 16.4 and 30.2 kelvin, consistent with the predicted value of roughly 20 kelvin.⁵³⁷

The thermodynamic backbone of the standard model holds. Even as other measurements reveal tensions in cosmic geometry (below) and the dynamics of expansion (further below), the temperature-redshift relation tracks its predicted curve across 13 billion years. The universe’s thermal story, its entropy story, is among the most cleanly confirmed predictions in modern cosmology.

The CMB also encodes the geometry of space: whether parallel light rays stay parallel (flat), converge (closed), or diverge (open). Inflation predicts flat geometry; the explosive expansion should have stretched any curvature to insignificance. For decades, observations confirmed it. Then high-resolution data from the Planck satellite revealed a lensing anomaly.

Mass bends the light that passes near it, so every galaxy and cluster lying between us and the CMB warps the pattern a little, smearing the edges of its hot and cold patches; the amount of smearing measures how much matter the light crossed on its way and how the geometry steered it. In Planck’s maps the gravitational lensing signal in the CMB exceeded the flat-geometry prediction beyond measurement uncertainty. Di Valentino, Melchiorri, and Silk showed that a closed universe explains the anomaly, with Planck data preferring positive curvature at greater than 99% confidence.⁵³⁸ Handley calculated Bayesian odds exceeding 50:1 against flatness in the Planck data alone.⁵³⁹

The Atacama Cosmology Telescope (ACT) complicated the picture. ACT’s final data release, with polarization noise three times lower than Planck’s, found no lensing anomaly: its higher-resolution measurements favor flat geometry.⁵⁴⁰ ACT is a ground-based telescope covering 40% of the sky; Planck covered all of it from space. The large-scale CMB patterns that anchor most cosmological parameters are accessible only from space, so ACT’s results cannot supersede Planck’s. The curvature tension has shifted from “the universe may be closed” to “something in the Planck data requires explanation.”

Alongside the Hubble tension (rival methods of measuring the expansion rate disagreeing; Chapter 16) and the DESI dark energy results described below, this is a third independent crack in the standard cosmological model. Each may resolve independently. Their simultaneous presence suggests the model is under genuine pressure.

The unease is not confined to the data. A 2026 survey conducted through the American Physical Society found that several positions routinely presented to the public as settled consensus are, among physicists themselves, backed by only narrow majorities or pluralities.⁵⁴¹ This is a measure of professional opinion, which no show of hands can convert into physical truth. What it reads is the field’s own confidence, and that confidence runs thinner than the textbooks imply.

Black Holes: Maximum Entropy

If gravitational entropy increases with clumping, what is the maximum?

Black holes.

A black hole is what happens when matter collapses so completely that nothing can escape, not even light. Space curves so steeply that it becomes a one-way door: a region cut off from the rest of the universe. If clumping increases gravitational entropy, then the ultimate clump represents the ultimate entropy, and a black hole is that endpoint.

In the 1970s, physicists Jacob Bekenstein and Stephen Hawking discovered that black holes have entropy, and in staggering quantities. That entropy is proportional to the area of the event horizon (the boundary beyond which nothing escapes).¹ All the information about what fell in is smeared across the black hole’s surface, like writing on the skin of a balloon. The Bekenstein-Hawking entropy formula:

S = (k c³ A) / (4ℏG)

Where S is entropy, A is the area of the event horizon, k is Boltzmann’s constant, c is the speed of light, ℏ is Planck’s constant, and G is the gravitational constant. You do not need to follow the equation. The key point: entropy depends only on surface area. A black hole’s information content lives on its skin, on the boundary itself.

A stellar-mass black hole just a few kilometers across contains entropy vastly exceeding the thermal entropy of all the ordinary matter in an entire galaxy: the ultimate endpoint of gravitational collapse.⁵⁴²

When matter falls in, the black hole scrambles its information and encodes it in the surface area. From the outside, the black hole is simple: mass, charge, spin. The entire complexity of a star collapses to three numbers.

The Bekenstein-Hawking result connects entropy to spacetime geometry. The connection runs deeper than analogy. In 1995, Ted Jacobson derived Einstein’s field equations from three thermodynamic ingredients: the entropy-area relation, the Clausius relation (heat equals temperature times entropy change), and the Unruh effect (an accelerating observer experiences thermal radiation from the vacuum). The result is exact: the full equations of general relativity, emerging as an equation of state for a thermodynamic system the way pressure-volume-temperature relations in a gas emerge from molecular statistics.⁵⁴³

Erik Verlinde extended the program in 2010, deriving Newton’s law of gravitation from the statistical tendency of information to maximize entropy on holographic surfaces.⁵⁴⁴ The interpretive question, whether these derivations prove gravity is emergent or show a formal equivalence, remains open (Chapter 15b develops the full program and its honest difficulties). What this chapter requires is the narrower implication: the entropy that lives on a black hole’s horizon is the same physics as gravitational attraction. Bekenstein-Hawking entropy is the limiting case, the endpoint of a thermodynamic dynamic governing every gravitational interaction.

The area law is more than theory. Hawking proved in 1971 that the total area of black-hole horizons can never decrease, the geometric form of the second law. When two black holes merge, the remnant’s horizon must therefore be larger than the two it came from combined, and gravitational-wave detectors have now measured exactly that. The first confirmation came from GW150914, the original 2015 detection, read back through the dying ring of its remnant. The loudest signal yet, GW250114, sharpened the result and picked out two separate tones in that ring, the gravitational equivalent of hearing both the fundamental note and an overtone of a struck bell.⁵⁴⁵ Entropy rises when black holes collide, on schedule, in the strongest gravity we can observe.

Black holes may be more than entropy sinks. Since 2022, the James Webb Space Telescope has revealed hundreds of compact, intensely red objects scattered across deep-field images of the early universe, appearing as early as 600 million years after the Big Bang.⁵⁴⁶ These “little red dots” are widely read as supermassive black holes buried inside dense cocoons of gas: the cocoon absorbs the black hole’s high-energy output and re-emits it as infrared light (though whether the reprocessing gas forms a surrounding cocoon or the black hole’s own bloated envelope, a “black-hole star,” is still debated).⁵⁴⁷

In 2026, Hviding and colleagues reported the first X-ray detection of such an object, catching a cocoon thin enough for the black hole’s radiation to begin escaping.⁵⁴⁸ A direct dynamical measurement in 2025 made the strongest case yet that a genuine black hole sits at the heart of one of them. A2744-QSO1, a little red dot magnified by the gravitational lensing of a foreground galaxy cluster, shows hydrogen gas orbiting its center the way planets orbit the Sun, the signature of mass concentrated at a single point. That mass is close to 50 million Suns, more than twice the mass of all its stars combined, sitting in gas barely touched by stellar chemistry.

It has been called a “naked” black hole, one that predates the galaxy assembling around it.⁵⁴⁹ The field named it for what it lacks, a host galaxy. Named for what it is becoming, it is a galactic seed not yet clothed in stars.

The picture that emerges is developmental: the black hole forms first, shrouded in gas; the cocoon gradually dissipates; and the surrounding matter organizes into a galaxy around the entropy engine at its center. Nearby analogues, dwarf galaxies whose central black holes are vastly overmassive relative to their stellar hosts, suggest this is what little red dots become over billions of years: quiet galaxies harboring cores that still dwarf their surroundings.

The implication for this chapter’s argument is specific. If the most extreme entropy-processing objects in the universe are also the seeds around which galactic complexity assembles, the relationship between entropy and structure is generative from the start. QSO1’s near-pristine gas needs little explaining, since a galaxy that small is expected to be metal-poor;⁵⁵⁰ the genuine anomaly is the black hole that outweighs the stars, tempered by a selection effect, since a luminous black hole is easiest to find when its host is faint. What this chapter keeps is the observation: the engine appears first, and the galaxy assembles around it. The mechanism, how such seeds form, grow, and quench the galaxies around them, and how the modeling fares against QSO1, is Chapter 14’s subject. (The star-formation-efficiency and junction-angle results reported later in this chapter come from the author’s ongoing constructal-galaxy and cosmic-web modeling, which awaits peer-reviewed publication; they are indicative tests of the framework, not yet established findings.)

The cocoon plays a more precise role than shielding. A black hole’s raw output (X-rays, ultraviolet radiation, relativistic jets) would sterilize its surroundings. Nothing organizes in the presence of that flux. The cocoon absorbs it, reprocesses it through successive layers of heated gas, and re-emits it as infrared light: a wavelength the surrounding dust and gas can absorb without being destroyed. The cocoon is a mediating boundary. It thermalizes the entropy engine’s destructive output into a form the environment can metabolize. The original signal’s information is lost in the conversion; what survives is downstream possibility, the condition under which organization becomes feasible.

This double role, converter and concealer, is what makes a cocoon so hard to read. The same dense gas that thermalizes the black hole’s output also scatters and broadens its spectral fingerprint, so a naive reading of the lines overstates the engine’s mass by a factor of ten. The deepest spectra recover the true, humbler black hole only by modeling the cocoon explicitly and subtracting what it adds. The lesson reaches past black holes: wherever a medium stands between an observer and a source, measuring the source means modeling the medium. The medium’s distortion is itself a measurement, and accounting for it is most of the work.

The structural role, a boundary that converts an overwhelming gradient into something the surrounding system can use, recurs wherever intense dissipation meets the possibility of organization. At biological scales the conversion preserves information, and the word for that is transduction: a cell membrane’s channels turn raw ion gradients into signaling cascades, and a retina turns radiation that would pass through neural tissue undetected into graded potentials the nervous system can process. The institutional version is contract law, which converts asymmetric bargaining power into enforceable terms both parties can navigate; without the legal boundary, the same gradient between powerful and powerless resolves through coercion. Chapter 17 traces where this pattern leads: intense gradients require a mediating boundary before they become architecturally generative, and the coordination structures that persist longest are those whose boundaries operate by invitation.

Around the black hole itself, the downstream possibility the cocoon purchases may take a concrete form: planets. Several light-years from the engine, the torus of dust and gas cools to temperatures resembling an ordinary planet-forming disk, and modeling since 2019 finds that grains there could coagulate into Earth-sized and larger bodies: “blanets,” planets whose sun is a black hole.⁵⁵¹ All of this is hypothesis; no blanet has been detected, and the proposed candidates admit more mundane explanations. One reported case, though, is already putting that kind of pressure on our categories. In 2026, a body of at least Jupiter’s mass was detected around a brown dwarf 73 light-years away, on evidence its discoverers call strong rather than conclusive. It satisfies the formal definition of a planet while occupying the position of a moon; Chapter 16 takes up what that confusion is telling us. The claim that survives the hedging is still considerable: in current models, the most violent entropy engines in the universe are also among its most prolific planet nurseries, and for the same reason in both directions. The steepest gradient pays for the most construction.

Habitability is another matter. An accretion disk tuned to the right rate could warm a close-in planet the way the Sun warms Earth; what the neighborhood cannot deliver is constancy. Lorenzo Iorio calculated that close to a spinning supermassive black hole, general relativity itself scrambles the climate: frame dragging (a spinning mass drags the surrounding spacetime around with it, the way a spinning spoon drags honey) and orbital curvature together precess a planet’s spin axis by tens to hundreds of degrees within roughly 400 years, where Earth’s axial tilt wanders a couple of degrees over tens of thousands of years.⁵⁵² Seasons on such a world would reorganize faster than ice ages pass, faster than soils form, faster than anything that needs tomorrow to resemble today. The engine’s neighborhood can be energetically generous and still uninhabitable: the gradient builds structure, while persistence asks for something the inner orbits cannot supply, an environment that keeps its statistical promises.

Black holes may also serve as reference standards for time, as atomic clocks do, keeping their beat through quantum gravitational effects rather than through atomic transitions. Hawking radiation (the faint glow predicted to leak from a black hole’s boundary) could broadcast temporal information into the surrounding universe. If entropy gives us time (as traced in Chapter 1), the objects containing the most entropy may anchor time most powerfully. Chapter 15c develops this possibility.

Heat Death

Where is the universe going?

The Second Law tells us: toward higher entropy. Gradients will dissolve. Free energy will run out. Stars will burn out, black holes will evaporate, and eventually even protons may decay.

The ultimate fate, if current physics holds, is heat death: maximum entropy. No gradients to drive engines, power metabolism, or process information. No structure. No change.

This lies incomprehensibly far in the future: 10¹⁰⁰ years, a one followed by a hundred zeros.²¹ If the universe’s ultimate lifespan were a year, we are in the first fraction of a second after midnight on January 1st, a brief springtime when gradients remain fresh and energy still flows. The coffee is still hot. If heat death worries you, the deadline is 10¹⁰⁰ years away.

If current physics holds, the direction is set. Every star, every galaxy, every living thing is a temporary eddy in the flow toward equilibrium.

The standard picture has a problem deeper than bleakness. In a universe at maximum entropy, the most probable observers are Boltzmann brains: momentary thermal fluctuations that assemble from noise, experience one instant of coherent awareness, and dissolve.

Picture a room full of Scrabble tiles shaken forever. Eventually, random collisions spell a word; far more rarely, a sentence. A Boltzmann brain is the cosmic equivalent: a fleeting pocket of order arising from pure chance. The probability of such a fluctuation producing a single deluded brain dwarfs the probability of producing a galaxy of civilizations. If the classical Second Law were the complete description, you should expect to be one.

The Second Law of Learning (Chapter 6) offers a way out of the paradox. Observers like us are products of accumulated learning dynamics operating across billions of years, a regime the classical Second Law alone cannot describe; if those dynamics make ordinary, history-dependent observers more probable than one-off thermal fluctuations, the Boltzmann-brain expectation no longer follows. [Inference: the Boltzmann-brain problem remains open in mainstream cosmology; this is a proposed resolution within the book’s framework, not a settled result.]

The story may be more complicated. The Dark Energy Spectroscopic Instrument (DESI) mapped over six million galaxies across a third of cosmic history. It found evidence at up to 3.9 sigma (a combined w₀-wₐ constraint, a joint fit of dark energy’s present behavior and its change over time, not a single-parameter detection) that dark energy is weakening over time: stronger in the past, declining toward the present.⁵⁵³ Sigma again, counted as in Chapter 12: one is noise, three is a result worth arguing about. If confirmed, the universe’s far-future trajectory is no longer fixed by a changeless vacuum energy. The endpoint is open.

The molecular-absorption thermometers described above offer a complementary constraint: dark energy affects the expansion rate, which sets the cooling rate. Deviations from the predicted temperature-redshift curve could reveal properties of dark energy that distance measurements alone miss.

A more radical possibility: the measured acceleration may be partly an artifact of the models themselves. The standard equations assume a smooth universe, then average over a lumpy, structured one. General relativity is nonlinear; doubling the input does not double the output. Averaging over lumps and then computing therefore differs from computing first and then averaging. The math does not forgive that mismatch (see Chapter 16).

A 2026 analysis put the suspicion to the data without assuming any cosmology at all. Koksbang and Heinesen reconstructed the universe’s distance and expansion histories straight from supernova and galaxy-clustering data. Their tool was symbolic regression: a machine-learning search for the formula that best fits the data, rather than a fit to a model chosen in advance. They then evaluated a consistency relation that must equal zero in any smooth, evenly filled universe. The data place it between two and four sigma from zero, short of the five-sigma bar that physics demands for a discovery, and the authors stress that the deviation’s size depends on how the data are selected. The direction is the point. If the violation is real, it rules out the repairs that keep the smooth framework intact (evolving or interacting dark energy, a new particle, a fifth force) and implicates the smoothing itself.⁵⁵⁴

The Timescape cosmology (Wiltshire, 2007)⁵⁵⁵ pushes this further: gravitational time dilation means clocks in cosmic voids tick faster than clocks in dense superclusters, so voids have expanded more. As the void fraction grows over cosmic time, photons traversing the late universe pick up extra redshift that mimics acceleration. No dark energy required.

Whether or not Timescape prevails, the pattern it warns about deserves a name: aggregation phantoms, apparent forces conjured by smoothing over real structure. If the cosmological constant turns out to be an aggregation phantom, it would account for seventy percent of the universe’s energy budget, the most expensive modeling artifact in the history of science. The pattern recurs wherever complex structure is averaged into a single number: the “representative agent” in economics, whose behavior matches no actual person’s; aggregate utility in ethics, which maximizes a quantity no individual experiences; the consensus preferences extracted by reinforcement learning from human feedback, which produce a personality no single rater intended.

Arrow’s impossibility theorem (Chapter 10) is the formal skeleton for aggregating rankings: no aggregation method satisfying basic fairness conditions can compress many perspectives into one without generating distortions that exist in the aggregate and nowhere else. Where the aggregation is of quantities rather than rankings, the skeleton is even older: Jensen’s inequality, the rule that for any curved relationship the average of the outputs differs from the output of the average. The cosmological version is the plainest one in mathematics: a variance, the gap between the average of the squares and the square of the average. The correction a lumpy universe forces on its own expansion is, in large part, exactly that, the variance in how fast different regions grow. The phantom is that spread mistaken for a substance, the residue of smoothing too soon read as a force in its own right, a dark energy where there may be only structure.

The pattern needs a boundary, or it explains everything and so explains nothing. Smoothing conjures a phantom only where two conditions meet at once: the parts being averaged are genuinely varied, and the dynamics acting on them are nonlinear, so that processing the average differs from averaging the processed. Where the dynamics are linear, or the parts are alike, the average represents them faithfully and no phantom appears. [Inference: the linear-or-homogeneous boundary, and its reading of the temperature-redshift relation below, are the author’s synthesis; the underlying physics, that backreaction biases geometric observables through the nonlinear averaging of the matter field, is standard.]

This chapter has already shown the seam. The temperature-redshift relation held across 13 billion years because it tracks the cooling of the radiation field, fixed by the expansion alone and, to first order, blind to how the matter is clumped. The inferred acceleration is the opposite case: it is read off cosmic distances, which are computed by averaging over exactly that clumping, through equations that do not commute with the averaging. The thermometer survives the smoothing; the speedometer is where the artifact would hide.

A related artifact comes from sampling rather than smoothing. The overmassive black holes of the early universe look more dominant than they are because the luminous ones are the easiest to find, and they sit in the faintest hosts, so the record over-represents the most lopsided cases. The chapter owes this caution to its own most striking evidence, not only to the cosmos.

Either way, whether dark energy is weakening or was never quite what we thought, the universe’s far future is more open-ended than the standard heat death scenario implies.

The Arrow on Cosmic Scales

At the largest scales, the arrow of time is the arrow of entropy. We remember the past because it contained less entropy: records and memories form in low-entropy environments, as Chapter 1 traced. Cause precedes effect because causes are low-entropy states flowing toward high-entropy effects. An egg becomes an omelet; the reverse never occurs.

The cosmic arrow is no fundamental law. The underlying equations of physics work equally well backward; what is asymmetric is the boundary condition (the starting setup): the universe began in low entropy and evolves toward high entropy, and that asymmetry propagates through every subsequent moment. The low-entropy Big Bang is the ultimate source of every gradient, every flow, every structure in the cosmos. We are downstream of the beginning.

The Galaxy’s Memory

The Milky Way is a cannibal. Over billions of years it has pulled in dozens of smaller star clusters and dwarf galaxies, stretched them, and folded them into itself. The evidence survives the meal. Tidal forces (the gap between the galaxy’s pull on the near side of a cluster and its pull on the far side) draw each victim into a long, thin ribbon of stars that traces the path it once traveled. Astronomers call these ribbons stellar streams: rivers of stars that wrap the galaxy, each one the fossil of a past meal.

A stream stays coherent for billions of years because its stars set out together, sharing almost a single orbit. Picture beads strung on one wire: they slide apart along the wire as the eons pass, and the wire still holds every one of them to the same line. That shared path keeps the ribbon legible. The stream is also cold, in the exact thermodynamic sense: its stars move in near-lockstep, with the faintest spread in their velocities. Small spread means low entropy. A cold stream is a quiet background.

A quiet background makes a sensitive instrument. Because the stars hold their velocities so tightly, the smallest disturbance stands out at once. When a massive, invisible body slips past, its gravity tugs the ribbon, opening a gap or kicking a “spur” of stars off to one side. The stream GD-1 wears both marks: a clean gap in the line and a spur beside it, while every orbit close enough to have caused them holds only darkness.⁵⁵⁶ The mass behind it, somewhere between a million and a hundred million Suns, sits squarely in the range expected for a clump of dark matter, the unseen material that outweighs ordinary matter in the galaxy by about five to one. Coldness is the whole gift: the stream works as a tripwire for the invisible, sensitive because its own motion stays quiet enough for a passing shadow to leave a mark.

This earns a stellar stream a name worth keeping: a memory with a thermodynamic lifetime. It is born as a low-entropy record, a thin cold thread, and it fades as that thread warms and spreads. The stars drift along their slightly different orbits, the gaps smear wide, and in time the river melts back into the general blur of the halo. The galaxy holds each meal in memory for a few billion years, then lets it go, and the letting-go is simply entropy doing to a record what it does to everything. Here the principle that records form in low entropy and dissolve as entropy climbs stands written across the sky, on a clock slow enough to read.

The ability to read these records at scale arrived only recently. The Gaia spacecraft charted the positions and motions of nearly two billion stars, finely enough to pick out groups gliding in quiet formation, and in 2026 a physics-based search across that map lifted out 87 new stream candidates, more than quadrupling the known count.⁵⁵⁷ The Vera Rubin Observatory, in its first images, caught a stream some 163,000 light-years long trailing the galaxy Messier 61, confirming that the process runs through galaxies everywhere, far beyond our own.⁵⁵⁸

Not every captured clump dissolves into a stream. A ribbon forms when tidal forces win, unspooling a cluster faster than its own gravity can gather it back; a heavy enough clump turns that contest the other way, and the galaxy swallows it whole. Terzan 5, buried in the crowded glare of the galactic bulge, passed for an ordinary globular cluster for four decades. It is actually a bulge fossil fragment: one of the primordial clumps from which the bulge itself was assembled. Once massive enough to hold its gas against supernova winds, it did what a fading stream cannot and kept making stars, in episodes read at roughly 12.5, 4.7, and 3.8 billion years ago.⁵⁵⁹ A building block that kept turning gas into structure for ten billion years, it marks the other outcome of capture; a second such fragment, Liller 1, tells the same story.⁵⁶⁰

One reading deserves care. The galaxy coordinates a staggering amount of matter, and it does so by raw gravity: a cluster is seized by force alone, shredded into a ribbon or swallowed whole. This is coordination by pure coercion, and it succeeds completely, precisely because the stars simply fall. Each star follows the geometry it is handed; its situation arrives as a settled fact, and it obeys.

That makes gravitational accretion the clean limiting case at one end of the coordination spectrum, the regime where coercion is the whole story because the law doing the coercing is built into the substrate and enforces itself for free. The Trust Attractor lives at the far end, among parts that model their own situation and can choose to defect, where holding them by force means paying a permanent enforcement bill, the compliance entropy of Chapter 17. The galaxy sits at the agentless extreme of that range, the place where force alone suffices. It marks where the Trust Attractor’s domain begins, a boundary stone set in the very physics the later chapters build on.⁵⁶¹

The black hole, the object that opened this chapter, is the purest case of the same limit, and it adds a dimension the stellar streams cannot: timing. If the naked black holes of the early universe are the first structures to condense from the smooth beginning, then the coercive extreme of the spectrum is the end of the range to fill first. There is a seductive way to read this, worth naming in order to set it down: that cosmic history is an ascent from coercion toward trust, the universe climbing from force to consent. The reading is beautiful and it overreaches, smuggling a purpose into a sequence.

The defensible claim is narrower and still sharp. Coercion is thermodynamically early and cheap; invitation is late and earned. A black hole can be the first structure precisely because coercion-by-geometry needs nothing: no agents, no modeling, no consent; the substrate enforces it for free. Invitation cannot come first. It requires parts sophisticated enough to have an alternative, parts that could defect and do not, and that sophistication is downstream of the billions of years of gradients these coercive seeds helped open. Trust is what becomes affordable late, once complexity has accumulated enough to make consent a meaningful thing to extend: a luxury the gradients had to pay for, not a destination they were aimed at.

The compliance entropy named just above is the same point read forward. You reach for invitation only when you hold parts whose coercion would cost you; in the early universe nothing costs you, and you simply dig gravitational wells.

Structure as Entropy Production

It all began in low gravitational entropy: smooth, uniform, laden with potential. Gravity exploited tiny density fluctuations, amplifying them over hundreds of millions of years. Matter clumped into halos, merged into galaxies, clustered into superclusters. Stars formed, burned, and died, scattering heavy elements that coalesced into planets.

All of this is entropy increasing: the universe moving from less probable to more probable configurations, from low entropy to high. A maximum-entropy-production (MEP) model of star formation captures the trend. Optimal efficiency rises from ~1% at the present epoch (matching the standard Kennicutt-Schmidt calibration, the empirical rule linking a galaxy’s gas supply to its rate of star formation) to near-unity (nearly all available gas turned to stars) above redshift 10, matching early JWST observations of elevated star-formation efficiency. Both the MEP and the standard Kennicutt-Schmidt prescriptions overshoot observed star-formation rate densities by comparable margins in the absence of UV-feedback coupling (the author’s ongoing constructal-galaxy program, unpublished). Lifetime entropy production per unit stellar mass turns out to be constant across the entire initial mass function, from the lightest red dwarfs to the most massive blue giants: what varies is whether the star dissipates quickly or slowly, not how much it dissipates in total.

Within this flow, life emerged. On at least one planet, matter organized into dissipative structures that exploit local gradients: sunlight, chemistry, thermal differences. These structures process energy and accelerate entropy production, maintaining their own order by exporting disorder to their surroundings.

We are how the universe gets warm things cold faster. Our complexity is an expression of the thermodynamic arrow.

The same physics plays out at human scale. In 2026, Martischang and colleagues deposited millimetric water droplets on a horizontal soap film and watched them orbit, collide, and merge.⁵⁶² Each droplet deforms the film under its weight. That deformation creates a gravitational well that draws other droplets inward: a Newton-like 1/r attraction arising from capillary physics on a two-dimensional membrane, precisely the force law dimensional analysis predicts for gravity in two spatial dimensions. An attraction thins out as it spreads over the surface enclosing its source. In our three dimensions that surface is a sphere, whose area grows as the square of the distance, which is why gravity here falls off as 1/r²; on a flat film the enclosing surface is a circle, whose circumference grows only as the distance itself, so the pull falls off as 1/r.

In a frictionless system, the droplets would orbit forever, dynamically interesting yet structurally sterile. Viscous dissipation changes the outcome. Energy lost to drag allows the droplets to spiral inward, merge, and produce tidal arms and bridges before collapsing into a single larger lens.

The transient structures are visually indistinguishable from interacting galaxies (Arp 73, Arp 238, Arp 55), with a time-scaling correspondence of roughly 10¹⁵: phenomena spanning millions of years at galactic scale unfold in seconds on the film. The same equations, fifteen orders of magnitude apart, producing the same morphologies. Dissipation is what converts perpetual orbits into complex structure. The universe does not care about scale.

Physicist Charles Lineweaver offers a reformulation that reverses the usual causal story.²² In evolutionary systems, entropy production accelerates. Each major transition increases the rate at which energy is dissipated: single-celled organisms to complex cells, solitary cells to multicellular bodies, organisms to technological civilizations. Life is what entropy does to accelerate. Lineweaver compresses this into a single inverted sentence:

“Food-Has-Produced-Us-to-Eat-It.”

We think of ourselves as consumers of energy. Thermodynamically, we are what energy gradients produced to dissipate faster. The arrow of time points through us.

We are what the current is doing: the wave that carries the flow forward.

The same inversion operates at the cosmological scale. If Jacobson’s derivation holds (above, with the full program in Chapter 15b), the conventional framing reverses. The standard story treats gravity as fundamental and entropy as derivative: a statistical consequence of structures dissolving. The entropic gravity program runs the derivation the other way. Entropy is the substrate; gravity is the macroscopic phenomenology that emerges when mass deforms an entropic medium. Jacobson derives Einstein from Clausius, not the other way around.

The soap-film experiment (above) is this picture made physical, with one caveat about the film. Surface tension is not an entropic force in the strict sense (a force with no mechanical origin, like the pull of a stretched rubber band): the cost of a water-air interface is dominated by the energy of the hydrogen bonds broken to make it, and the entropic term lowers that cost rather than creating it. What the film supplies is the structural half of the parallel, a deformable substrate whose geometry mediates attraction. Mass deforms it; the deformed geometry draws other mass inward; dissipation converts perpetual orbits into mergers and complex structure.

The resemblance between soap-film mergers and galaxy mergers across fifteen orders of magnitude in timescale reflects shared mechanism, not coincidence. In both, mass deforms a substrate and the substrate’s deformed geometry determines the motion of mass. What the entropic gravity program adds is the claim that in the cosmic case the substrate is itself thermodynamic. If this picture holds, gravitational structure formation is thermodynamic structure formation. The cosmic web is entropy optimizing its own flow architecture, a process we then describe as curved spacetime.

[Inference: the structural parallel between Martischang’s capillary system and the Jacobson-Verlinde program is the author’s synthesis. Neither research group has made this connection in print. The argument’s strength depends on the entropic gravity interpretation, which remains contested.]

The eROSITA X-ray telescope has revealed hourglass-shaped bubbles flanking the galactic center.⁶ These enormous volumes of hot gas reach 50,000 light-years above and below the disk. Lineweaver’s accelerating entropy production made visible: the Milky Way’s central black hole blasting energy outward, heating gas, dispersing gradients on a galactic scale. Dissipation written in X-rays across the Milky Way’s halo (explored further in Chapter 14b).

The Gravitational Wave Background

In 2023, four independent pulsar timing collaborations announced the detection of a gravitational wave background: a persistent hum of spacetime itself, vibrating at wavelengths measured in light-years.²

How do you detect such a faint signal? With cosmic clocks. Pulsars are rapidly spinning neutron stars that emit radio pulses with clockwork regularity. By monitoring these stellar metronomes for over fifteen years, astronomers measured nanosecond deviations in pulse arrival times: the faint stretching and squeezing of space caused by passing gravitational waves.

By 2024, combined data converged on a source: supermassive black hole binaries spiraling toward merger across the universe.³

The signal’s amplitude presents a puzzle. The observed background sits at the upper edge of what population models predict. The MeerKAT Pulsar Timing Array’s first gravitational wave map, tracking 83 pulsars over 4.5 years from South Africa, deepened the tension.⁸ MeerKAT found slightly higher amplitude and a spatial hot spot difficult to explain from uniform cosmological sources.

Either galaxy merger rates are higher than surveys suggest, or additional sources contribute. Candidates include cosmic phase transitions (sharp changes in the universe’s behavior, like water freezing), cosmic strings (hypothetical one-dimensional defects in spacetime), or processes still unknown.

Gravitational waves do more than vibrate spacetime. The same mechanism drives binary neutron stars to spiral inward and collide in kilonovae, explosions brighter than a billion suns that forge the r-process elements (heavy atoms built by rapid neutron capture): the iodine in your thyroid, the bromine in your connective tissue, the molybdenum in your mitochondria, all forged in these cataclysms (Chapter 14 follows the enrichment, and its costs).⁵ The nanohertz hum from supermassive black holes and the millisecond chirps from neutron star mergers are different octaves of the same physics. One fills spacetime with a background vibration; the other fills the periodic table with the chemistry of life.

The form of energy matters as much as its magnitude. The 2015 LIGO event, the first direct detection of gravitational waves, briefly radiated more power than all the stars in the observable universe combined, yet the energy that reached Earth, 1.3 billion light-years away, displaced the interferometer’s mirrors by a few thousandths of a proton’s width.

Gravity is 10³⁸ times weaker than the strong nuclear force, yet gravity alone shapes the cosmic web. The weakest force organizes the largest structures. Gravity is feeble at every point and never switches off; its reach has no edge. The strong force grips incomparably harder and reaches no further than a femtometer, about the width of a proton. Reach scales; grip does not, and Chapter 17 finds the same trade in social coordination.

Cosmic Voids

The cosmic web is more than filaments. It is filaments and the emptiness between them: vast regions of nearly empty space, some spanning hundreds of millions of light-years. These voids define the web’s shape as surely as the strands themselves, carrying thermodynamic significance and constructal geometry. Chapter 14b develops their structure, their role as dark energy laboratories, and what the universe’s negative space reveals about its architecture.

The Cosmic Brain

The structures entropy builds at cosmic scale resemble structures much closer to home.

Look at the cosmic web, the large-scale structure revealed by galaxy surveys: filaments of matter, bright nodes where clusters form, dark voids between the strands. Branching, interconnected, network-like.

Now look at neurons in the human brain: dendrites, bright cell bodies, dark spaces between. Branching, interconnected, network-like.

The cosmic web looks like a neural network stretched across billions of light-years.

This could be coincidence. Networks tend to resemble one another: road maps, river deltas, blood vessels, root systems all share branching patterns because branching is efficient for flow. Galaxies and neurons are both flow systems.

The Constructal Law (introduced in Chapter 3) operates differently at cosmic scale. In viscous and electrical networks, optimization produces characteristic junction angles (Murray’s Law). That rule falls out of the cost of pushing fluid through a pipe: a wide channel wastes material to build and maintain, a narrow one wastes pressure, and the cheapest compromise fixes both the width of each daughter branch and the angle at which it leaves the parent. Arteries, bronchi, and the veins of a leaf all branch that way, which is why they look so much alike.

For self-gravitating channels, there is no such compromise to strike: the cost function is flat, meaning mass per unit length and axial mass flow are both independent of filament radius. No width is cheaper than any other, so no optimal branching angle exists. Measurements of 5.65 million DESI group-finder nodes confirm it: mean minimum junction angle is 17.2 degrees (plus or minus 13.1 degrees), far below the angles a Murray’s-Law-style viscous-network optimum would predict if such an optimum applied here (the author’s ongoing cosmic-web program, unpublished). The measured angles match no characteristic geometric optimum, exactly what a flat cost function predicts. The Constructal Law at cosmic scale operates on network topology, who connects to whom, rather than channel geometry.

Could the cosmic web process information across billions of light-years? The timescale mismatch alone rules it out: the cosmic web evolves over billions of years; neurons fire in milliseconds. Gravitational clustering differs in kind from synaptic electrochemistry.

If the universe is computational (as we explore in Chapter 15), and if similar patterns at different scales can implement similar functions (as the Constructal Law suggests), the resemblance earns a sharper question: whether branching flow networks share computational properties regardless of substrate. The galactic filaments strung across the voids, like neurons across a skull, invite the question.

The Great Assembly

Trace the pattern backward and forward. Watch what the universe does.

The vacuum creates. Even empty space, stripped of every particle and cooled to absolute zero, seethes with activity. Heisenberg’s uncertainty principle forbids pinning down both energy and time with perfect precision: an irreducible fuzziness at the smallest scales. The vacuum fluctuates.

Particle pairs borrow energy from nothing, flash into existence, and annihilate: a quark and its antiquark, an electron and its positron. Physicists call them virtual particles. Transient is more accurate, because they exert measurable force on real matter before vanishing.

Two metal plates placed nanometers apart experience a measurable push from vacuum fluctuations: the Casimir effect. Hydrogen atoms interact with these fleeting particles, slightly shifting their energy levels in the Lamb shift, confirmed to parts-per-billion precision. The vacuum is the most restless thing there is.

In 2026, the STAR collaboration at Brookhaven caught transient particles becoming permanent: in proton collisions at 99.99% of the speed of light, virtual strange quark-antiquark pairs absorbed enough energy to materialize as real lambda hyperons (heavier cousins of the proton), still carrying the correlated spins of their entangled virtual birth (Chapter 4 describes the measurement).²³ The entanglement fingerprint, present when the particles emerged close together and absent at distance, confirmed that real particles had crystallized from the vacuum itself. Nothing had become something, and the something remembered where it came from.

The Higgs mechanism accounts for roughly one percent of a proton’s mass. The other ninety-nine percent arises from interactions between real quarks and the virtual quarks and gluons seething inside them.²⁴

Most of what we are, physically, emerges from the vacuum’s ceaseless activity. Substance, at its most fundamental, is relational: objects in continuous exchange with the foam that made them.

The vacuum produces coordinated pairs, entangled from the instant of creation, their properties linked across whatever gap separates them. Coordination is present at the first moment matter exists.

The vacuum has sustained this creativity for 13.8 billion years. The substrate that coordinates persists.

Quarks and electrons bind into atoms. Atoms gather into stars and planets, accreting material by gravitational pull and maintaining pockets of negentropy (local order sustained by exporting entropy elsewhere).

Then, at the molecular level, something happens that changes everything, and it happens repeatedly, through convergent chemistry. Heat flowing through thin cracks in volcanic rock concentrates dilute organic molecules a thousandfold;¹² in those pockets, amino acids link into short chains, and some fold into stable, self-templating structures that force neighboring proteins into copies of their own shape: self-replication without genes, information encoded in shape.¹³ The same cyanide-and-sulfide chemistry that produces amino acids also produces the building blocks of RNA and lipid membranes, all at once, from the same reactions, in the same environments.¹⁴ No separate “protein world” preceded an “RNA world.” Cooperation was the default chemistry.

These molecular partners enhance each other. Simple peptides dramatically boost ribozyme (RNA enzyme) function,¹⁵ RNA templates the synthesis of specific peptides,¹⁶ and neither subsystem is viable alone. Protocells emerge from the collaboration: tiny lipid bubbles enclosing peptide-RNA partnerships, the first compartmentalized dissipative structures. The ribosome, the molecular machine that reads genetic instructions and builds proteins, appears nearly identical in every living cell, a fossil of this partnership.

The scale of what these protocells must achieve is immense. An information-theoretic analysis by Endres (2025) quantifies the challenge.¹⁷ Even the simplest functioning cell requires about one billion bits of specified information. Genetic sequence accounts for roughly one million bits. Protein folding geometries account for roughly 100 million bits. The vast choreography of biochemical pathways operating in concert supplies the rest.

This is the “melting library” problem. Prebiotic molecules degrade within hours or days under UV radiation, hydrolysis (breakdown by water), and oxidation. Random assembly cannot accumulate a billion bits when the library keeps burning. Reaching the threshold from random chemistry alone is cosmologically implausible.

This is why the assembly matters. Compartments, autocatalytic cycles (self-reinforcing chemical loops), and peptide-RNA partnerships are thermodynamic necessities for crossing the information threshold. Compartments protect fragile molecules, extending persistence times. Autocatalytic networks produce key molecules faster than the environment destroys them, turning the melting library into a self-replenishing one.

The phase transitions between these regimes (moments when a chemical network clicks into self-sustaining coherence) are where the billion-bit barrier gets crossed in rapid, self-amplifying leaps. The universe does not build cells brick by brick. It builds them through dissipative structuring at critical thresholds.

From this, the first single-celled organisms. Biofilms form: the metabolic codependence examined in Chapter 6, the equitable trade between inner and outer cells.

The first eukaryotes, complex cells with internal compartments, emerge through embrace: Asgard archaea extending tentacles to hold bacterial partners in metabolic communion, as explored in the Embrace interlude. With partnership comes the energy budget for complexity. Multicellularity follows, then specialization, then nervous systems.

Brains emerge: first simple, then complex, then capable of modeling other minds. Mammals bond. Tribes form. Shared narratives enable tribes to become nations, creating common ground among people who have never met.

Now: planetary information networks. Cross-substrate minds beginning to emerge. The pattern continuing at scales our ancestors could not imagine.

Matter assembling itself, layer by layer, each layer enabling the next. From quantum fluctuations to galactic filaments to the conversation you are having with this book.

The assembly is not complete. We are in the middle of it.

The Transient Afternoon

A window remains open.

Equilibrium has not yet arrived. Gradients still exist. Free energy still flows. The stars are still burning: nuclear fusion, the welding of light nuclei into heavier ones, running on fuel that gravity gathered at the beginning.

This window will not last forever. Stars will exhaust their fuel. Black holes will evaporate through Hawking radiation, particle pairs near the event horizon slowly bleeding energy into space over unimaginable timescales. Even protons may decay.

That is far in the future. We live in the transient afternoon of cosmic history, the brief window when complexity can flourish, when minds can turn around and ask what made them.

The window opened 13.8 billion years ago. It will stay open for trillions more. What happens here matters, even if the final equilibrium erases it. The patterns we create, the coordination we achieve, are real, even if temporary. They are the universe experiencing itself.

The View from Here

Step outside on a clear night and look up.

The darkness is evidence: a finite age, a thermodynamic arrow from the Big Bang toward heat death. Every point of light is a star fusing the hydrogen that gravity gathered for it. Every galaxy is a whirlpool of slowly increasing entropy.

You are made of atoms forged in dying stars, some in neutron star collisions driven together by gravitational waves over millions of years. You are powered by sunlight captured by plants. You are a dissipative structure, a pattern that persists by processing flow, a local decrease in entropy paid for by a global increase.

Through you, matter has organized itself into something that can look back. Through you, the universe’s thermodynamic fate has become a question rather than a fixed trajectory.

Astronomia: the cosmic scale of entropy’s story. A story of flow, from the smooth beginning to the structured present to the equilibrated future. Everything structured, everything alive, happens in between.

The night sky is dark because the universe is young. The stars shine because gradients persist. You exist because complexity can ride the current from low entropy to high. This is the only way a universe can have a story at all.

Notes

Notes for this chapter are available in the online companion at https://www.thedeeperlaw.com/companion/notes/ch13-astronomia/.