Order That Repeats vs. Information That Specifies
Origins · · 11 min read · By No Design Editorial
Energy flowing through a system makes pattern: hurricanes, convection cells, snowflakes. It has never been shown to make code. That gap is the whole argument.
Every design argument that appeals to order runs into the same reply, usually within a minute, and the reply is a good one. The Earth is not a sealed box. Energy pours in from the sun and radiates back out to space, and in a system like that, order arises on its own, at no cost, with nobody arranging it. Snowflakes form. Crystals grow. Heat a shallow dish of oil from below and it organises itself into a neat lattice of cells. Appealing to entropy against any of this is a category error, because the second law of thermodynamics forbids nothing of the kind in an open system.
That objection is correct, and it is worth conceding completely before going any further. Not partially, not with a qualification held back for later. It is correct, and a design case that leans on a misreading of thermodynamics deserves to lose, because a good deal of design writing has leaned on exactly that and has been rightly embarrassed for it.
But notice how narrow the objection is once you have granted it. It establishes that energy flowing through a system produces structure. The question is whether it produces the other thing.
Order does arrive for free, and here is the receipt
Take the cleanest demonstration in the physics. Put a shallow layer of fluid between two plates and heat it from below. Below a certain threshold nothing visible happens and heat crosses by conduction. Above the threshold the fluid abruptly organises itself into rolls or hexagonal cells, and heat starts moving by convection. This is Rayleigh–Bénard convection, and the threshold is a single dimensionless number, the Rayleigh number, which for a layer held between rigid plates sits at about 1,708. What makes it striking is that the figure does not depend on the fluid. Water, silicone oil, air. The threshold belongs to the geometry and the driving, not to the chemistry.
The Belousov–Zhabotinsky reaction does the same trick in time rather than space: a beaker of inorganic chemistry that oscillates in colour and throws out rotating spiral waves, running through thousands of cycles with no template and no instructions. Ilya Prigogine won the Nobel Prize in Chemistry in 1977 for the theory that covers all of this, the theory of dissipative structures, and it is a real and beautiful piece of science.
It is also worth quoting Prigogine on what he thought he had shown. The formation and maintenance of self-organising systems, he and his colleagues wrote in 1972, are "compatible with the laws of physical chemistry." Compatible with. Not produced by, not explained by. That is a careful sentence from a man who had every incentive to write a bolder one.
Now turn the heat off. The convection cells decay within seconds and the fluid returns to a featureless layer. Seal the Belousov–Zhabotinsky flask and the oscillations run down and stop when the reagents are spent. Nothing is retained. These structures are not stored anywhere; they are a shape the energy flux is holding in place, and they last precisely as long as the flux does. No cell remembers the cell before it. No pattern is copied. Nothing is handed on.
Pattern is not the thing that needs explaining
A hurricane is one of the most organised objects on the planet. It has a defined eye, a symmetric eyewall, spiral rainbands, and it runs as a genuine heat engine converting a temperature difference into motion. It is also entirely uninformative. It stores nothing, specifies nothing, makes no copy of itself, and leaves no description behind. Sunlight falling on an ocean produces hurricanes reliably, year after year. It has never once produced an instruction.
That is the line the whole argument runs along, and it is worth stating in one sentence. Energy makes pattern. The question is whether it makes code.
What a snowflake actually is
The snowflake is the example everyone reaches for, so it is worth being precise about it rather than romantic.
Kenneth Libbrecht at Caltech has spent decades growing snow crystals under controlled conditions, and his account is unglamorous. The six-fold symmetry comes from the shape of the water molecule, which favours a hexagonal lattice. The elaborate branching comes from the interplay of surface attachment kinetics and vapour diffusion. And the six arms resemble each other, when they do, because "the individual branches all experience nearly the same changing conditions as a function of time." They are not signalling to one another. They are correlated because they shared a journey through the same weather.
Libbrecht is also blunt that "the vast majority of natural snow crystals are lopsided or malformed to varying degrees." The flawless six-pointed stars are a photogenic minority, selected by photographers who do not publish the blobs.
So the complexity of a snowflake is real, and it is generated fresh every time by local growth rules plus a shared trip through changing air. Nothing is stored, nothing is copied, nothing is read, and nothing is passed to the next crystal. Two snowflakes forming a metre apart are not related in any sense; they are two independent records of similar weather. And nobody has measured a snowflake's information content in bits, whatever the figures circulating online suggest. The comparison with a genome was never a comparison of quantities. It is a comparison of kinds.
"Non-repeating" is not enough, and here is why
There is a sharper version of the objection that deserves a proper answer, because the word "non-repeating" cannot carry the argument on its own.
In 1984 Dan Shechtman found a metallic alloy with five-fold symmetry that crystallography had held to be impossible, and took the 2011 Nobel Prize in Chemistry for it. These are quasicrystals: long-range order that never repeats. And in 1988 Onoda, Steinhardt, DiVincenzo and Socolar showed something that ought to give any design writer pause, demonstrating that defect-free quasiperiodic tilings "can be constructed by use of local rules alone."
So a non-repeating structure with long-range order can be grown from short-range interactions with no blueprint anywhere in the system. If the design argument were "structure that never repeats requires a mind," it would end here, and it should.
It is not that argument. Look at what a Penrose tiling cannot do. Every finite patch of tiles in one Penrose tiling occurs, infinitely often, in every other Penrose tiling. There are uncountably many distinct tilings and no local region will ever tell you which one you are looking at. Mathematicians call this local isomorphism, and it means the tiling cannot carry a locally readable message, because every legal local arrangement appears everywhere, at the same frequency.
A genome is the precise inverse. Its local patches are exactly what distinguish one sequence from every other, and any of the possible sequences is roughly as stable as any other, which is what leaves the sequence free to be arbitrary. Aperiodicity is a geometric property. Information capacity requires something else: arrangements that are locally distinguishable, energetically near-equivalent, and actually read.
Erwin Schrödinger, who coined the phrase "aperiodic crystal" in 1944 for the then-unknown carrier of heredity, chose it because periodicity would exclude carrying information. The International Union of Crystallography, which now owns the term technically, notes drily that the relationship between his phrase and real quasicrystals "is not functional, merely semantic."
Why the letters are free
Here is the fact at the centre of the whole question, and it is a fact about chemistry, not philosophy.
The bond attaching a base to the sugar-phosphate backbone of DNA has effectively the same energy whichever of the four bases it is. Put an A where a G was and the molecule is no less stable, no more strained, no less content. Chemistry has no preference about the order of the letters.
Michael Polanyi put the consequence as sharply as it has ever been put, in the journal Science in 1968. "As the arrangement of a printed page is extraneous to the chemistry of the printed page," he wrote, "so is the base sequence in a DNA molecule extraneous to the chemical forces at work in the DNA molecule. It is this physical indeterminacy of the sequence that produces the improbability of any particular sequence and thereby enables it to have a meaning."
That is the point people miss when they call this a gap in our knowledge. It is not a gap. It is a precondition. If chemistry did fix the order, DNA would be a repeating crystal, and a repeating crystal can no more spell a sentence than a magnet holding iron filings in a fixed pattern can write one. The indifference of the chemistry is what makes the message possible.
Leslie Orgel, a founding figure in origin-of-life chemistry and no friend of religious argument, drew the same line in 1973 and gave it its name. "Living organisms are distinguished by their specified complexity," he wrote. "Crystals are usually taken as the prototypes of simple well-specified structures, because they consist of a very large number of identical molecules packed together in a uniform way. Lumps of granite or random mixtures of polymers are examples of structures that are complex but not specified. The crystals fail to qualify as living because they lack complexity; the mixtures of polymers fail to qualify because they lack specificity."
Two notes on that passage. The truncated version in circulation, which has granite as a crystal, is corrupted and incoherent, and quoting it will cost you the argument in front of anyone who has read the book. And Orgel was describing what he believed natural selection produces, not proposing a test for design. Borrowing his phrase does not enlist him.
A code is only a code to something that reads it
Now the part that the entropy objection never reaches.
In 1948, five years before anyone knew the structure of DNA, John von Neumann worked out what a machine would need in order to reproduce itself in an open-ended way. His answer was an architecture, not a chemistry. You need a constructor that builds from a description, a copier that duplicates the description without understanding it, a controller that sequences the two, and the description itself. The decisive feature is that the description gets used twice, in two incompatible modes: read as instructions, and copied blindly as data. Copy the machine directly and you cannot accumulate heritable change. Interpret the description without copying it and you cannot pass it on.
Biology turned out to be built exactly that way. DNA is translated and DNA is replicated, by separate machinery, and von Neumann derived the requirement before anyone could have told him.
The tighter knot is in the code itself. The mapping from codon to amino acid is not performed by base pairing. It is performed by twenty aminoacyl-tRNA synthetase enzymes, each of which recognises one amino acid and the transfer RNAs that carry the matching anticodons, and joins them. Those enzymes are the reader. And every one of them is written in the language it reads, and can only be built by the machinery it makes possible. Three researchers stated the problem plainly in the journal Transcription in 2018: "there is no known mechanism to generate aaRS proteins until the code has evolved," and, of their own proposal, "at this time, we offer no simple solution to this problem." Charles Carter and Peter Wills, who have spent years building a naturalistic model of exactly this bootstrap, put the difficulty in one line: "blind selection cannot read phenotypic properties and copy them back into genetic messages."
That is why turning energy into coded information requires a system that already reads the code. It is not a probability estimate. It is a structural fact about what a code is.
Four things the design side should stop saying
An argument is worth more when it names its own bad currency, so here are four claims that should be dropped.
Information cannot increase by natural processes. This is false, and it is demonstrably false. Richard Lenski's long-running E. coli experiment produced a population able to use citrate aerobically, and the 2012 genomic analysis traced it to a tandem duplication that placed an existing citrate transporter under a promoter that fires in oxygen. Joakim Näsvall and colleagues ran the innovation-amplification-divergence route to completion in Salmonella in the laboratory the same year. New genes arise. Say so.
DNA carries vast Shannon information, therefore design. Shannon himself ruled this out on the first page of his 1948 paper: "these semantic aspects of communication are irrelevant to the engineering problem." A string of random noise maximises Shannon information. Quantity of information was never the argument.
Specified complexity is a rigorous detector of design. It is not, at least not yet. The formal apparatus has been taken apart in the philosophical literature, notably by Elsberry and Shallit in Synthese in 2011 and by Olle Häggström on the No Free Lunch theorems, and the central practical objection is unanswered: computing a specification requires knowing the space of alternatives and their probabilities, which for a biological structure nobody knows.
The codon assignments are entirely arbitrary. Not entirely. Michael Yarus's work on RNA aptamers found statistically significant binding affinities between several amino acids and their own cognate triplets, six of the eight amino acids tested in his 2017 review. The effect is real. Eugene Koonin's assessment is that it is nowhere near sufficient to account for the code, partly because the amino acids showing it are ones that probably arrived late. The accurate statement is that the mapping is largely, but not wholly, physically unconstrained, and the residue is what does the work.
Notice what all of the successful demonstrations have in common. Every one of them happens inside a coded system that already exists. Citrate use is promoter capture within a working genome. None of this literature addresses where the code came from, and the researchers involved do not claim it does.
The strongest naturalistic attempt, stated fairly
The most serious current attempt to get life-like organisation out of energy flux is Jeremy England's work on dissipative adaptation, beginning with a 2013 paper in the Journal of Chemical Physics. It derives a bound: the faster and more durable a self-replicator is, the more heat it must dump into its surroundings. Applied to E. coli, the measured heat output runs about five to six times the absolute physical minimum, which is a genuinely interesting result. Later work produced chemical networks that settle into rare high-dissipation states more often than chance, and a swarm of vibrating robots that self-organises into low-agitation configurations.
What none of it produced is a heritable sequence, a symbol-to-referent mapping, or anything that reads. England says so himself. "I'm not saying this tells me a lot about what's going on in a biological system," he told Quanta in 2017, "nor even claiming that this is necessarily telling us where life as we know it came from." Sara Imari Walker's verdict is that the theory is necessary but not sufficient, because it does not account for life's information processing.
Stuart Kauffman's autocatalytic sets, the "order for free" programme, are the other serious candidate, and here the honest number comes from a paper Kauffman co-authored in 2012: a viable autocatalytic core amounts to "one bit of heritable information." That is a real result and it is not nothing. It is also about nine orders of magnitude short of a bacterial genome.
One further counterexample deserves a hearing, because it is the best one available. In evolutionary game theory, arbitrary signalling conventions do emerge spontaneously: Brian Skyrms and others have shown that a sender and a receiver with no agreed vocabulary converge on an arbitrary signal-to-state mapping through reinforcement alone. That is, formally, a code appearing without a designer. But look at what the model is handed at the start: an agent with internal states, an agent with a repertoire of actions, a channel between them, and a shared payoff for getting it right. Given all that, the particular vocabulary comes free. The models show how a code gets fixed once translation exists. They do not deliver the sender, the receiver or the channel, and those are the things in question.
What is left standing
Strip all of this down and the position is straightforward, and it is narrower than the design argument is usually made to be.
The entropy objection is not answered by denying it. It is answered by pointing out that it was aimed at a claim this book does not make. Yes, the Earth is open. Yes, energy flux builds structure. Convection cells, spiral chemical waves, snow crystals, quasicrystals that never repeat: all of it real, all of it unaided, none of it in dispute.
And none of it is a message. No experiment has produced a symbolic code from undirected chemistry, and the people closest to the problem say so in print. Paul Davies, who is looking hard for a physical rather than a theological answer, frames the origin of life as the question of "how encoded software emerged spontaneously from hardware," and adds that he does not believe the known laws of physics can explain it. Eugene Koonin's summary of half a century of work on the origin of the genetic code is that the question "might remain pertinent even in another fifty years."
That is a gap in the explanation, not a proof of impossibility, and it should be described as exactly that. But it is a very particular gap, and it has not moved in seventy years of trying. Energy has never been shown to write anything, because writing is not a thing energy does. Writing is what a mind does with energy.
'He created everything and determined it with precise determination.' (Quran 25:2)
From the Introduction to No Design.
Read the rest of the argument at No Design, and find the book and the wider project at How to Muslim.
Tags: Information, Thermodynamics, Genetic Code, Quran