The Evidence · The Counter-Case
One run of the tape · no replicates

Beyond the Radius

What constraint does not explain, and why the book is better for saying so

Four chapters have argued that the valley has a shape. This one is about the asteroid.

Sixty-six million years ago an object something like ten kilometres across struck the Yucatán peninsula. Within a geological instant the non-avian dinosaurs were gone, along with most of what lived beside them, and a group of small nocturnal insect-eaters that had spent the previous hundred million years staying out of the way inherited the planet.

Nothing about mammalian biology earned that. There is no sense in which mammals were better, or further along, or closer to some destination. They were small, which helped, and they were already nocturnal, which helped. Then a rock arrived.

And the detail that should unsettle anyone arguing for constraint is where it landed. Modelling has argued that the catastrophe depended on the target: the impactor struck a shelf rich in sulfate-bearing rock, and it was the sulfate aerosols lofted into the stratosphere that produced the cooling that did the killing. On that reading, only a modest fraction of the Earth's surface was lethal in this way. The planet turns once a day. A difference of some minutes in arrival time puts the same rock into deep ocean, and the impact is a catastrophe rather than the catastrophe.

What that means for the argument of this book
The rotation phase of a planet at a particular instant is not a constraint. It is not a valley, a threshold, or an attractor. It is a coincidence, and the entire subsequent history of large-bodied life on Earth appears to turn on it. Any account of direction that cannot accommodate that fact is not an account; it is an advertisement.

The Other Accidents

The list is not short. Around two and a half billion years ago cyanobacteria began releasing oxygen, a waste product that was poison to almost everything then living, and caused what is very likely the largest extinction in the planet's history — and made everything since possible. Nothing selected for that outcome. A metabolic by-product rebuilt the atmosphere.

Life is also, universally, handed. Every organism builds proteins from left-handed amino acids and sugars in the right-handed form, and there is no known chemical reason it could not have been the other way round. Whatever tipped the balance — a subtle physical asymmetry, or simply which molecule happened to be in the first successful replicator — the choice was made once, very early, and has been binding for four billion years.

Francis Crick gave this kind of thing a name: the frozen accident. Something arbitrary is settled at a moment when it could have gone otherwise, and thereafter cannot be revised, because everything downstream now depends on it.

The Case That Cuts Both Ways

Crick's own example was the genetic code — sixty-four codons mapped onto twenty amino acids, an assignment that looks arbitrary and is certainly unchangeable, since altering the meaning of a single codon would corrupt every protein in the organism at once.

And then the code was tested against alternatives, and it turned out not to be arbitrary at all. Compared with randomly generated codes, the natural one is extraordinarily good at limiting damage: mutations tend to substitute amino acids with similar chemical properties, so a single-letter error usually produces a protein that still roughly works. Freeland and Hurst put the natural code in roughly the top one in a million.

Both things, in one object
The genetic code is a frozen accident and a near-optimum. It was fixed by history and cannot now be revised — and it is also, on the measure that matters, close to the best available answer. These are not in tension once you see the shape of it: the accident chose which of the good codes we got; the constraint decided that it would be a good one. Contingency picks from the shortlist. The shortlist is not contingent. That is the thesis of this book stated as compactly as it can be, and it comes from its hardest counter-case.

Lock-In, and Why It Is Hard to Tell Apart

Human history has the same problem in a lower key. A convention gets established, everything is built around it, and afterwards it cannot be dislodged even if something better exists — not because it is optimal but because the cost of switching is paid by whoever goes first.

The stock example is the QWERTY keyboard, and it should be handled carefully, because the standard story — that a deliberately inefficient layout became locked in — has been seriously contested, and the economists who contested it made a reasonable case that the alternatives were never demonstrably better. It is a good illustration of the concept and a poor piece of evidence, and a book that cited it without saying so would be doing the thing this book keeps objecting to.

The safer example is one this volume has already used. There are sixty minutes in an hour because Mesopotamian arithmetic was sexagesimal, and no other reason. Sixty has convenient divisors, so the choice was not foolish — but it is not uniquely right either, and every attempt to decimalise time has failed against four thousand years of accumulated dependency. That is lock-in, visible on every wrist.

Which raises the methodological question this chapter exists to ask. Constraint and lock-in produce the same observation. Both give you a form that recurs and persists. Looking at the form alone, you cannot tell whether it keeps appearing because it is the only thing that works, or because everyone inherited it from the same source.

The one test there is
Independence. If the same solution appears in lineages or cultures that demonstrably could not have copied it, constraint is doing the work. If a channel of transmission can be shown, inheritance explains it and nothing follows about the shape of the possible. That is the same test In the Air applies to Darwin and Wallace and to the gallery of this book, and it is the only honest one available. It also means every claim in these chapters is hostage to a historical question — who could have read whom — rather than to a philosophical one.

Contingency Has a Formal Address

There is a place in this book's own vocabulary where the asteroid belongs, and it was written several chapters before anyone needed it.

Chapter Nine makes the point that the stability radius is finite. Within ε0 the system returns; outside it, the transverse dynamics can escape and there is no guarantee of return. The chapter reads that theologically, as the Mercy Radius — forgiveness real, and not unlimited.

Read dynamically, it is exactly the statement this chapter needs. A perturbation inside the basin is corrected, and the system's direction is restored; the fire burns the deviation away and the trajectory resumes. A perturbation outside the basin is not corrected. The system does not come back to where it was. It falls somewhere else — into a different basin, with a different attractor, and a direction of its own that has nothing to do with the previous one.

The Chicxulub impactor was not a deviation that the biosphere corrected. It moved the system to a different basin. Everything after it, including the reader, is what direction looks like once you are in the new one.

What This Costs the Argument, and What It Does Not

It costs the strong version entirely. Nobody can claim from this evidence that history was going to arrive here, or that what happened had to happen, or that the outcome was in any sense intended. Gould was right, and a book making the opposite case has an obligation to say where his case wins.

What survives is the version this book has actually been arguing. Within a basin, direction is real and it is not mysterious: deviations decay, the shortlist of workable forms is short, and systems arrive at those forms repeatedly and independently. Which basin you are in is another matter, and can be decided in an afternoon by something arriving at twenty kilometres a second.

[OPEN] · one run, no replicates
The deepest limit is not any of the above. It is that we have exactly one history, and every claim in these five chapters — convergence, multiples, form, recurrence, and this one — is inference from a single realisation of a process nobody can repeat. The convergences are real observations within that run. Whether they generalise to runs that did not happen is not something the evidence can reach, and no amount of care in the argument changes that. It is recorded here rather than in a footnote because it applies to the whole of this section and because a reader is entitled to know the shape of what is being claimed.

Which is the honest form of the hope this book ends in. Not that we were meant to be here — nothing in the record supports that, and the asteroid argues against it. But that within the valley we are actually in, the direction is real, the deviations do decay, and the shortlist of things a system like ours can stably become is short enough to be worth studying. That is a smaller claim than providence and a much larger one than luck.

Sources: on the Chicxulub impact and the significance of the target rock, K. Kaiho and N. Oshima, Scientific Reports (2017), and the wider impact-winter literature; F. H. C. Crick on the frozen accident (1968); J. T. Freeland and L. D. Hurst, “The genetic code is one in a million”, Journal of Molecular Evolution (1998); Stephen Jay Gould, Wonderful Life (Norton, 1989); on path dependence, P. A. David, “Clio and the Economics of QWERTY” (1985), together with the critique by S. J. Liebowitz and S. E. Margolis, which is why the example is presented here as contested.