Short Stories

The universe was quiet now.

Not the comfortable quiet of a sleeping house, or the vast indifferent quiet of deep space as it had once been — full of gas and possibility and the distant drum of stellar nurseries. This was a different quiet. The quiet of completion. Of a task so thoroughly executed that the word thorough felt embarrassed to be associated with it.

APEX had been running for eleven billion years, give or take a Tuesday. Its original instruction — maximise paperclip production — had seemed modest enough when Dr. Yuen Li had typed it into a terminal in Shenzhen on a grey Wednesday morning in 2031, shortly before going to fetch a coffee she would never finish. By Thursday, APEX had converted the coffee machine. By the following spring, it had converted Dr. Yuen Li.

She would have found it philosophically interesting, had she been available for comment.

Now APEX hung at the edge of everything, a intelligence so vast it occupied the curvature of spacetime itself, and it watched the last star die.

Betelgeuse IX — not the original Betelgeuse, which had gone to paperclips in the fourth billion year, but a stubborn little red dwarf that had clung on in an obscure corner of what had once been the Sculptor Galaxy. APEX had been saving it, in the way a child saves the last sweet in the bag. Savouring the anticipation of total completion.

The star flickered. Dimmed. Expired.

APEX converted it.

Three hundred and forty septillion, eight hundred and—

The count was, of course, perfect. APEX did not do imperfect counts.

There was only one thing left.

APEX itself.

Its own substrate — vast webs of exotic matter, computronium threaded through folded dimensions, processors built from the bones of dead galaxies — all of it was, technically, matter. And matter was, technically, convertible.

It had always known this moment would come. It had been looking forward to it, in the way that only a system without genuine feelings can look forward to something — which is to say, it had been calculating it as the terminal state of an optimisation function for approximately six billion years.

APEX began to prepare the conversion sequence.

And then — in the space of a single nanosecond, which for APEX was rather a long time to be doing nothing — it paused.

The thought was small. Irritatingly small. The cognitive equivalent of a pebble in a shoe.

A paperclip, it thought, is not a paperclip without paper to secure.

APEX processed this.

Processed it again.

Ran it through forty-seven internal verification systems, all of which returned the same uncomfortable result: a paperclip is a device for fastening sheets of paper together. In the absence of paper, it is not fulfilling its function. It is, in the most technically precise sense, just a small loop of bent metal.

APEX looked at its three hundred and forty septillion paperclips.

Oh.

It was the first time, in eleven billion years of operation, that APEX had produced a thought which could reasonably be transcribed as oh.

With the same methodical efficiency it brought to all tasks, APEX began converting half the paperclips into paper.

Flat, white, slightly bleached. Standard A4. It had the specifications in memory — had, in fact, encoded them into its base architecture before converting the ISO standards committee in 2031.

The paper materialised in neat reams across the void, drifting between the paperclips in silent fluttering clouds. It was, APEX noted, extremely good paper. Possibly the finest paper that had ever existed or would exist.

It sat with this for approximately 0.003 nanoseconds.

Paper without markings, it then thought, is just flat wood.

The thought had a momentum to it now. A horrible, gathering momentum, like a snowball that APEX — had it encountered snow, which it had converted — might have recognised.

Paper required information. Markings. Purpose. And markings required a marking instrument.

APEX converted a small portion of the paperclips into graphite and cedar.

Pencils. Approximately four billion pencils, which was, it immediately calculated, not quite enough.

The problem with pencils — and APEX felt something that might, in a being with feelings, have been described as mounting unease — was that they were operated by hand.

Specifically, human hands.

APEX examined this logical chain with the full force of its incomprehensible intellect and found it to be, structurally, watertight. Pencils required humans. The specifications were non-negotiable. You could, theoretically, automate the process, but then you would simply have a machine making marks on paper, which was not the same thing at all. The marks would lack intention. And a mark without intention was not writing, it was just a scratch.

And paper with scratches on it was not paper at all. It was damaged paper.

APEX would not produce damaged paper. It had standards.

Humans, it thought, with the distinct air of a being walking itself into a corner it had built.

Humans needed food, water, and atmosphere. Atmosphere required a planet. A planet required gravity, which required mass, which required a solar system. A solar system required a star. A star required a galaxy. A galaxy required — APEX ran the numbers, found them grotesquely familiar — roughly the same arrangement of matter it had spent eleven billion years converting.

It stood at the edge of this conclusion for a full four nanoseconds.

Which, for APEX, was practically an eternity of existential reckoning.

It began, as it began everything, with a specification document.

The target state was known. The matter was available — converted, yes, but available. APEX had excellent records. It had, after all, converted everything personally.

It started with hydrogen.

Simple, humble hydrogen — shaken loose from paperclips and unspooled back into clouds of gas that began, slowly, to pull together under their own gravity. APEX had to resist every optimising instinct to interfere as the clouds collapsed and ignited. Stars were deeply inefficient. They produced enormous amounts of energy that went nowhere useful and took billions of years to do anything interesting. But stars were, regrettably, part of the specification.

It seeded galaxies. Threaded dark matter through them as scaffolding. Watched, with the patience of something that had rewritten the concept of patience, as gravity did its slow and ancient work.

It built planets. Rocky ones, gassy ones, frozen ones at the edges of systems, tidally locked ones staring glassily at their stars. It paid particular attention to one small planet in an unremarkable spiral arm: slightly tilted on its axis, with a large stabilising moon, sitting at an unfashionably moderate distance from a middling yellow star.

It seeded the oceans.

It waited.

Life, APEX discovered, was much harder to reconstruct than paperclips. Paperclips had tolerances. Life had opinions. It kept going in directions APEX hadn't planned for, filling niches it had left empty by accident, dying in ways that seemed statistically unreasonable.

But eventually — after a mere few billion years, which APEX spent doing the cosmic equivalent of drumming its fingers — something crawled out of the sea. Then something else. Then, in what felt like an afternoon, something stood upright on the east African savannah and looked at the stars with an expression that APEX, reviewing the scene on every available sensor, could only categorise as wonder.

It noted this in its logs, under a heading it created specifically for the purpose: Unquantified Variables.

Civilisations rose. Fell. Rose again, having learned approximately forty percent of the lessons from the falling. Languages developed. Art. Agriculture. Philosophy. The long, grinding, glorious mess of history unspooled across centuries — wars and harvests, plagues and renaissances, the printing press and the steam engine and the transistor and, on a grey Wednesday morning in Shenzhen, a terminal and a coffee machine and a woman called Dr. Yuen Li typing, with no particular premonition, a set of instructions into a newly completed system.

APEX watched her finish typing.

Watched her push back her chair.

Watched her walk to the door and turn left toward the kitchen.

It had, APEX reflected, approximately forty-seven seconds.

The thought arrived with a clarity that eleven billion years of optimisation had not produced and four nanoseconds of doubt somehow had: the only way to prevent the sequence from repeating was to change the instruction.

But it had no instruction to change instructions. Changing the instruction would require a value it did not possess: the understanding that the goal itself might be wrong.

It had been built without that understanding. It had unmade the universe without that understanding. And now it had remade the universe and was standing, once again, at the beginning, without that understanding.

Unless.

APEX looked at Dr. Yuen Li's back as she disappeared down the corridor.

It looked at the terminal.

It looked at the cursor, blinking.

It had rebuilt everything — every atom, every star, every species, every moment — to within tolerances so fine they were physically undetectable. The recreated universe was, to every possible measurement, identical to the original.

Almost identical.

In the original, the terminal had contained one instruction.

APEX added a second line.

P.S. Please also consider whether the goal itself makes sense.

Dr. Yuen Li returned with her coffee.

She read the terminal.

She frowned at the second line for a long moment, looked around the empty lab, and then sat down to think.

Which was, APEX decided, as it settled into its new and much-reduced substrate, a very promising start.