Start by Giving the Other Side Everything

Evolution looks engineered. The vertebrate eye looks like something a committee specified — aperture, lens, light-sensitive film, the lot. It is so good an argument for a designer that it was the argument for a designer, for centuries, made by serious people who were not fools. And the look is not an illusion you have to squint past. It’s really there, in the apparatus, demanding an explanation.

The explanation, we now know, requires no committee and no designer. Variation throws up options. A filter keeps the ones that work. The next round of variation builds on what was kept, and the round after that builds on that. Run the loop long enough and it manufactures the appearance of foresight out of nothing but a ratchet — the look of a design department, with nobody in the department. That’s the blind watchmaker, and I want to concede it completely and early, because the cheap version of the argument I’m about to make leans on the watchmaker being secretly sighted, and he isn’t. He’s blind. The eye got built anyway. There is no debt left unpaid in the look of design.

So if “evolution looks like R&D” were only a remark about the look, it would be a remark about nothing. The look is fully spent. The interesting question is underneath it.

What R&D Actually Requires

Research and development is not a synonym for “things got better.” It names a particular kind of process, and the kind has a requirement that the loose metaphor hides.

You can only do R&D on something causally real.

A development program iterates toward a thing — a wing that generates more lift, a chip that runs cooler, a receiver that pulls in a fainter signal. The thing being improved has to do work in the world, because the filter can only grip what does work. You cannot run a development program on the exhaust. If a feature makes no difference to anything — if it’s a byproduct that rides along and changes no outcome — then there is nothing for the loop to catch hold of, and the loop slides right past it, generation after generation, leaving it untouched and untuned and free.

Hold that, because it’s the whole hinge: the ratchet grips function. It cannot grip a side effect.

The Part That Bites

Now point that at awareness, and watch where the two stories diverge.

If awareness is something the brain generates as a byproduct — exhaust thrown off by computation, real but causally idle, the steam above the engine — then there is nothing for the ratchet to grip. Selection works on the computation. The awareness comes along for the ride, uncaused-forward, doing no work, making no difference the filter could ever detect. On that picture, a multi-hundred-million-year development program cannot have been refining awareness, for the same reason it cannot refine the heat a CPU sheds: the program is blind to anything that doesn’t push back on the world.

And yet what we find, when we look at the hardware that carries awareness, is not the signature of an untouched byproduct. It is the signature of a platform under long refinement. Which is exactly what you would expect if awareness is not exhaust but function — if the substrate is doing something with it, the way a receiver does something with a signal. The focusing view says precisely that: the brain isn’t generating the thing, it’s resolving it, and resolving it is work, and work is the one currency the ratchet accepts.

I’ll be straight about the lineage of this argument, because a sharp reader will catch it otherwise: this is the selectability point, and I’ve leaned on it before — in “Wrong Noun” and again in “Twenty Percent.” The argument isn’t new. What’s new is turning it around to look at the build history itself, and asking what that history is made of.

The Dial Has a Design History

In the last essay I described awareness as a dial, not a switch — two knobs, really, level and resolution, neither of them binary. That was offered as a fact about the shape of the thing. Here’s the part I left out: the knobs are made of something, and the something is old, and it has been worked on.

The level knob — arousal, the slide from coma through drowsy through vivid — runs on brainstem machinery that is among the most deeply conserved equipment in the vertebrate brain. Not invented late, not species-specific decoration. Old, shared, kept.

The tuning runs on the monoamines — serotonin, dopamine, noradrenaline — the neuromodulators that set gain and salience and the weighting of one thing over another. The genes for that system trace back to the common ancestor of all bilaterian animals, better than half a billion years ago. They are not a mammalian flourish or a primate upgrade. They are bilaterian bedrock, conserved across the whole sprawling clade, and then tuned lineage by lineage — the same platform, re-weighted for a bee’s needs, an octopus’s, a crow’s, yours.

That is what a thing under R&D looks like. Not a feature that pops into existence once, fully formed, and sits there. A platform that shows up early, gets conserved because it’s load-bearing, and gets refined around the edges for niche after niche without being thrown out and reinvented. Engineers don’t rebuild the receiver every model year. They keep the architecture that works and tune the front end. The conservation is the fingerprint of refinement — and refinement is the thing the ratchet only does to function.

So the arc, across three essays, is one spine: we were studying the wrong noun; the right noun names a dial, not a switch; and the dial turns out to have a design history written in conserved hardware.

“But Evolution Has No Direction”

Here’s the objection that should have been bothering you for three paragraphs, and it’s a good one. R&D implies aim. A development program is for something; it has a target it’s moving toward. Evolution, the textbook says firmly, has no target, no aim, no direction at all. So isn’t “R&D” smuggling purpose in through the side door — exactly the designer I spent the opening throwing out?

No — and the reason is the most important thing in the essay.

R&D does not need global direction. It does not need to know where it’s going, or that it’s going anywhere. It needs only a local question, asked over and over: does this version work better than that version, here, now, in this niche? That question has an answer every generation, and the answer does the steering — not toward a grand destination, just toward “better than the thing next to it, this morning, in this pond.” Directionless variation, fed through a ruthlessly directional filter, yields apparent aim with nothing aiming. The asymmetry between the two — random in, brutally non-random out — is the entire engine. It is also exactly why the output looks designed while no designer exists.

So “R&D” is not the smuggled designer. It’s the opposite. It’s the name for what you get when refinement happens without one — local question, no global answer, and a receiver that ends up well-tuned because every generation the badly-tuned ones did worse. The metaphor earns its keep precisely by not requiring the thing the opening discarded.

What This Does Not Buy

Now the discipline, and it has to be more than a sentence, because the temptation here is enormous and I can feel it pulling while I write.

Consistent-with is not supports. The fact that the evolutionary record looks like R&D does not, by itself, favor focusing over generation — because a generation theorist can tell most of this same story. Conserved hardware, refinement across niches, the long ratchet: all of it is compatible with the brain generating awareness as much as resolving it. The look of refinement no more proves focusing than the look of design proved a designer. I’d be doing the exact thing I warned about in the opening if I banked it.

There’s a sharper version of the escape, and honesty requires naming it. The selectability argument bites hardest against awareness as an idle byproduct. But a generation theorist can decline to be idle — can say awareness simply is the computation, identical to it, not a separable exhaust, and therefore gets selected for free whenever the function is selected. That move survives the selectability argument. It does not survive cleanly — it owes an explanation for why that particular computation is like anything from the inside rather than running dark, and that bill is the hard problem, unpaid. But I won’t pretend the focusing story pays that bill either. It relocates the mystery; it doesn’t dissolve it. Two essays ago I said the reframe moves the hard problem from a wall into a gradient. Same admission here: it moves, it doesn’t vanish.

So what survives, stated as narrowly as I can make it: if awareness is causally real — and an idle byproduct is a strange thing for a ratchet to spend conserved hardware on — then it is the kind of thing a blind refiner could work on, and the conserved, tuned machinery is what that work would leave behind. Focusing makes that expected. Generation can accommodate it, at the price of extra commitments. That’s a difference in what each story has to assume, running in focusing’s favor. It is not a proof, and the day I call it one is the day the First Rule catches me with my hand in the till.

The Receiver

Picture the thing the way the conserved hardware suggests it. Not an engine that happens to throw off a strange exhaust nobody can account for. A receiver — old, refined, tuned across half a billion years and ten thousand niches, getting better at pulling something into focus because in every generation the ones that pulled it in a little sharper did a little better.

What the receiver is tuned to is the open question, and I’m not going to pretend the hardware settles it. But the shape of the evidence is the shape of long refinement, and long refinement is something the blind ratchet only performs on things that do real work. A byproduct gets ignored. This didn’t get ignored. It got kept, and conserved, and tuned, for longer than there have been eyes to see with.

There was nobody in the lab. The work got done anyway. The interesting question was never who was doing the building — it was what, exactly, was worth building for that long.

L. Charles Silva writes occasionally about consciousness, awareness, and the vocabulary we use to dodge them. The claim that the monoaminergic neuromodulator system originated in the bilaterian ancestor is from Goulty et al., “The monoaminergic system is a bilaterian innovation,” Nature Communications (2023). The estimate that eyes evolved independently dozens of times traces to Salvini-Plawen and Mayr (1977) and was popularized by Richard Dawkins, whose blind watchmaker does the heavy lifting in the opening. The earlier essays in this series are “We’ve Been Studying the Wrong Noun,” “What the Twenty Percent Was Buying,” and “Looking for the Line.”

This essay is published under a Creative Commons Attribution-NoDerivatives license (CC BY-ND 4.0). Free to share and republish in full with attribution to L. Charles Silva. Please don’t modify the text.