The Room With More Doors

or, why blind luck stalls at a horizon it can never cross, and a process with memory keeps finding rooms that shouldn't exist yet

Jump to the simulation: race blind luck against a process that keeps what works and builds new things out of old ones, and watch one of them acquire a direction

Here is a fact so ordinary we forget to be astonished by it: the world keeps getting more elaborate, and it does so faster and faster. It took life something like two billion years to get from a bare cell to a cell with a nucleus, and a few thousand to get from the wheel to the aircraft to the machine that is arranging these words. Something changed in the rate. And underneath the change in rate is a change in kind that this site has mostly danced around: most of the emergence in these pages has no direction at all. A flock is not going anywhere. A sandpile is not building toward anything. Evolution, the great engine of the living world, is famously blind — it has no memory of yesterday's good idea and no picture of tomorrow's, it simply keeps whatever survives this afternoon. And yet somehow, on top of all that aimless machinery, a few processes have appeared that do seem to head somewhere: that accumulate, that build on themselves, that reach places the blind version would wait out the heat-death of the universe to stumble on. This page is about how that happens, and about being honest concerning what it does and doesn't promise.

Start with the blindness, because it is the default and it is worth feeling how limited it is. Imagine you are searching for good designs by pure luck. Each try is independent: you reach into the space of possible things, pull out a random one, and see if it works. Reach near where you started and you'll turn up plenty — the easy stuff, the low-hanging combinations, is thick on the ground. But the genuinely far-out designs, the ones that would take a hundred lucky choices to line up all at once, are exponentially rarer the farther out you go. Pure luck will fill in the neighborhood around your starting point beautifully and then simply stop, camped at a horizon it can approach but effectively never cross, because crossing it would take a coincidence too large to ever come up. No memory, no accumulation, no direction. Just a puddle of the easy, spreading until it can spread no more.

Now add two ingredients — the same two, it turns out, whether you are talking about brains, tools, science, or life itself — and the puddle turns into an arrow.

The experiment below runs both at once so you can watch the difference open up. To one side is a map of what has been made — every dot a design, and the farther a dot sits from the corner, the farther past the starting point it lies. Blind luck plays in cool blue; the cumulative process, the one with the ratchet and the recombination, plays in gold. Beside it is a running tally of how many things each process has produced over time. Turn it on and give it a few seconds. You are watching a direction appear out of a rule that contains no direction anywhere in it.

Things to try:

Two different lessons are on the screen at once, and they are worth pulling apart because they are genuinely different claims. The map is about reach, and reach is where the word direction earns its keep. Blue fills in the easy neighborhood and then parks at its horizon, exactly as pure luck must. Gold does not park. It builds a rung, then stands on that rung to build the next, and the frontier of what it can reach marches outward without any single step being large or lucky — each new thing is only a little beyond the things it was made from. Given time it will reach any distance at all, into regions blind luck would never visit in the lifetime of a star. That outward march is the closest thing to a direction the universe gives you for free. Nobody aimed it. It falls out of nothing more than keep what works and combine what you have.

The chart is about rate, and it is the one that has a famous name. Blue's count crawls along the floor: each blind attempt is as likely to pay off as the last, so its production is roughly a flat trickle forever. Gold's count bends upward and keeps bending, because the number of new things it can try is set by how much it has already made — more parts, more combinations, more still to build. Growth that feeds on its own output is growth that accelerates. This is what Ray Kurzweil named the Law of Accelerating Returns: not merely that technology improves, but that the rate of improvement itself climbs, because each generation of tools is the raw material for the next. It is the exponential curve of the Lily Pad seen from underneath — not just the shape of the growth, but the reason there is a shape at all.

Ray Kurzweil built a career on pushing that observation to its edge, and it is worth following him out there before we pull back. His claim is not merely that a given technology improves exponentially, but that the exponential itself steepens — because progress is a relay of paradigms, each handed off to the next as it exhausts itself. Mechanical calculators gave way to relays, relays to vacuum tubes, tubes to transistors, transistors to integrated circuits; plot computation per dollar across all five and you get not one S-curve but a staircase of them, and the staircase as a whole bends upward faster than any single step — an exponential of exponentials. In The Singularity Is Near (2005) he ran that curve forward to a startling horizon: a moment, a few decades out, when the pace of change grows so steep that the future stops being legible from where we stand. You need not buy the date, or the destination, to feel the pull of the engine underneath it — which is exactly what the gold curve draws for you, live, on the chart: acceleration that comes from no single clever trick but from the pile of tricks itself paying for the next one. The technologist Kevin Kelly gave that accumulating whole a name, the technium, and argued it has wants in precisely this site's sense of the word — not desires but reliable tendencies, and this outward-and-upward pull is the chief of them. It is where this site got the word in its title.

The room with more doors

There is a lovely image for the thing gold is doing, and it comes from the biologist Stuart Kauffman by way of the writer Steven Johnson. Kauffman called the set of things reachable in one step from where you are now the adjacent possible — not everything imaginable, just everything one recombination away from what already exists. The house of the possible, in Johnson's telling, is a strange one: you start in a few rooms, and each room you enter reveals doors you couldn't have seen from the last one, opening onto rooms that were not reachable until you arrived. “The strange and beautiful truth about the adjacent possible,” Johnson wrote, “is that its boundaries grow as you explore them.” Every door you open adds new doors. That is precisely what the gold process is doing on the map: each design it makes is itself a new ingredient, a new door, and so the space of what it can reach next is always expanding faster than it is being used up. Blind luck stands in the entrance hall trying random keys on the outer wall. The cumulative process walks the palace one room at a time, and the palace keeps getting bigger the deeper it goes.

It is worth being precise about how the adjacent possible relates to the luck horizon you can see on the map, because they look like the same boundary and are in fact opposites. The luck horizon is a wall that never moves: a fixed radius, set once and for all by the sheer improbability of far coincidences, that blind luck approaches and can never push past. The adjacent possible is the other kind of edge entirely — a frontier that recedes as you reach it, because every design you complete becomes a new ingredient and opens doors that were not there a moment ago. Same starting corner, two utterly different fates: luck is boxed in by a horizon that stays put; the cumulative process rides a horizon that runs from it, always one room further out than it was. That receding edge is the whole reason gold never has to stop where blue does.

And here the picture is telling a mercy of a lie that is worth catching it in, because the truth is even more in gold's favor. The map is two-dimensional; the real space is not, not remotely. Flattening everything that has been made onto a little quarter-plane is a cartoonist's kindness — the actual space of the possible has an axis for every domain there is: one for metallurgy, one for optics, one for organic chemistry, one for number theory, one for the ways to move heat, one for the handling of information, one for each of a thousand others, and a design is a single point carrying a coordinate on all of them at once. That is not a pedantic correction; it is the reason the adjacent possible is as explosive as it is. In two dimensions a point has only a handful of near neighbors. In a space of thousands of dimensions it has an astronomical number of directions in which to take its next step — and the richest steps are the ones that reach across axes: a trick from chemistry married to a trick from machining, a scrap of pure math that turns out to describe a piece of physics, a printing press that is really a wine press that met a coin punch. The flat map keeps the lesson legible, and that is its job; just keep in the back of your mind that the room it stands for has not four walls but thousands, and that the doors joining far-apart domains are exactly where the largest undiscovered rooms tend to hide.

Stack those two facts — that the adjacent possible widens as you explore it, and that it does so in a space of thousands of dimensions rather than two — and you arrive at the thing Kurzweil, Peter Diamandis, and Kevin Kelly all keep circling: not merely exponential growth, but growth whose exponent is itself rising. The doublings are getting shorter. Whether that meta-rate climbs linearly or exponentially is something people argue about, and, as you put it, it hardly matters — a shrinking doubling time is a shrinking doubling time. The mechanism is right there on the map. More parts make more combinations; but more domains make more kinds of combination, so it is not only that the fuel grows as it burns — the rate at which the fuel grows is growing too. Diamandis reads that curve as the case for abundance: energy, food, information, and medicine getting radically cheaper as the combinations compound, scarcity dissolving faster than our intuitions were built to track. He is at least partly right, and it is genuinely thrilling. It is also, in the same breath, a little terrifying — because a world that reorganizes itself faster than the institutions and the hard-won wisdom meant to steer it can keep up is a world that grows more powerful without automatically growing wiser. That gap is the whole subject of the next section. Nothing on this screen decides whether the acceleration carries us into the abundance or over a cliff; that part, as ever, was never in the ingredients.

Notice, if you flip the two switches under the experiment, that the two ingredients do genuinely different jobs — you can take either one away and watch which lesson dies. Turn recombination off, and the cumulative process keeps its ratchet: it still builds on what it has, still out-reaches blind luck, still heads somewhere. But its count stops accelerating and settles into a plain straight climb, because now each innovation is a lone invention rather than a marriage of everything that came before — the fuel no longer grows as it burns. Reach survives; acceleration dies. The ratchet is what gives the process a direction; recombination is what makes that direction speed up. You need both to get the full hockey stick, and pulling them apart on the screen is the cleanest way I know to feel that they are two ideas, not one.

And this, finally, is where Kelly's phrasing for the whole site cuts hardest, because it clarifies what is and isn't special about us. Blind biological evolution built everything alive, and it did it with the ratchet alone — genes are a memory, faithfully copied, so good tricks stay won — but with almost no recombination across lineages (your genes and an oak tree's never meet) and, crucially, with no foresight anywhere. It optimizes for whatever survives this afternoon and nothing else. What arrived late, and what changed the rate, is a kind of accumulation that can also borrow sideways and aim: culture, language, science, technology, and now the machines we are building — systems that pool their innovations across every mind at once, and that can, sometimes, point. That is the sense in which a physicist once said we may be the only way the universe has of understanding itself: not because we are outside the process, but because in us the blind ratchet grew a second ingredient and, with it, the beginnings of an aim. It is a real difference. It is also, as the next section insists, not a promise about where the aim will point or how far it can go.

Honesty about the arrow

It would be easy to leave the experiment glowing with the gold curve heading for the sky, and easy is exactly the problem, because the unbounded version is a fantasy and everyone selling it knows better. So turn the Ceiling on and watch what actually happens to any real accelerating process. Give the space an edge — a finite pool of usable combinations, a physics that won't be cheated, a market that saturates — and the gold hockey stick does not climb forever. It accelerates, magnificently, and then it bends over and flattens into an S, because the doors run out: the adjacent possible is real but it is not infinite in any given room, and as the room fills, the fraction of new tries that land somewhere genuinely new falls toward zero. This is the site's own recurring caution, met here in its home territory. Every exponential you have ever been shown is a logistic that had not yet reached its ceiling. Moore's Law, the poster child for accelerating returns, was a stack of S-curves handed off one to the next, each new device technology rescuing the trend as the last one flattened — and there is no law of the universe guaranteeing that the next rescuer always arrives on time. Acceleration is a tendency, not a promise.

There is a subtler honesty owed here too, about the word direction, because it is doing lighter work than it sounds like. The gold process heads outward, reliably, into the far reaches of its space — but nothing in the machinery says the far reaches are better, only that they are farther, more elaborate, harder to have reached by luck. “Farther from the starting point” is not a moral coordinate. A cumulative process with a ratchet and a knack for recombination will just as happily build its way toward weapons, surveillance, addictive machines, and brittle over-optimized systems as toward medicine and music; the arrow points away from the simple, not toward the good. And the ratchet, the very thing that gives the process its memory, is also what makes its mistakes so hard to undo — a bad standard, a bad institution, a bad dependency gets frozen in and built upon exactly as faithfully as a good one, which is the dark twin of this whole story and a page of its own. This is the spine the site keeps returning to and will not soften even here, on the most optimistic page in the collection: recurrence is not endorsement, and neither is acceleration. That a process reliably speeds up and heads somewhere tells you it is powerful. It does not tell you the somewhere is worth arriving at, and deciding that is the one part of the job no ratchet will ever do for us.

Still — and the honesty is what earns the “still” — something genuinely new does sit on that screen, and it is worth ending on how strange it is. Take a rule with no goal in it anywhere: keep what works, and build the new out of the old. Run it next to pure luck, which is what most of the universe has to work with. Luck fills its little neighborhood and stops, forever, at a horizon set by the sheer improbability of coincidence. The other one doesn't stop. It builds a floor to stand on and reaches higher, and the reaching makes more floor, and the more floor there is the faster it can reach — until, with no one steering and nothing aimed, a frontier walks out across the whole space and keeps accelerating into rooms that had no doors until it arrived to open them. It is not that the universe wants to head anywhere; it wants nothing, in the only sense this site ever means the word — the arrow is simply what these two plain ingredients reliably tend to produce when a slow world happens to stumble on both at once. It took life two billion years to find the first ingredient and a few hundred thousand to find the second. We are what happened next, and the direction we are riding is not a law of nature promising to carry us somewhere good. It is a tendency we were handed and now have to steer.