initial commit: the night trilogy (1988, 2026, 2088)

This commit is contained in:
Hermes Cron Job
2026-06-20 10:32:09 +00:00
commit 786c1a0084
3 changed files with 131 additions and 0 deletions
+43
View File
@@ -0,0 +1,43 @@
# The Night Before the Machines Could See
London, June 1988.
The computer on the desk had sixty-four kilobytes of RAM and cost more than the television. It could display sixteen colours at once, which was considered obscene. The manual called it *"a personal computing revolution."* Martin called it *"the thing that won't print."*
He was twenty-four, which in 1988 meant you were old enough to have bought a C64 as a teenager and young enough to still believe that the next machine you bought would actually be the last one you needed. It never was.
Outside, the rain was doing that English thing where it doesn't so much fall as hover, uncertain, like it's waiting for permission to commit. Martin's flat smelled of damp plaster and the burnt toast that had been his breakfast, and on the television, the Argentina versus England match was in the sixtieth minute. Nil-nil. It had been nil-nil since the first minute, which was either the most boring football match in history or the most tense, depending on whether you were Argentine or English. For Martin, who was English but had started liking Maradona somewhere between the hand of God and his own loss of faith in authority, it was both.
The machine hummed. Not metaphorically — it actually hummed, a high-pitched whine from the power supply that Martin's landlord had called *"a health hazard"* and Martin's flatmate had called *"the sound of the future"* before moving out because of the humming. The future, it turned out, sounded like a refrigerator full of wasps.
On screen, a BASIC program was running — a crude neural network simulation Martin had cobbled together from a *Scientific American* article and a book from the library he'd had to renew three times. It was supposed to learn patterns. What it actually did was get confidently wrong about everything, which Martin found more interesting than if it had been right. It looked at a training set of twenty football results and concluded that the colour of the away team's socks was the primary predictor of outcome. *"Fascinating,"* he'd written in his notebook, and then spent an evening wondering if that was what all intelligence was: pattern-matching with delusions of significance.
The telephone rang. His mother, from Colchester, telling him he'd missed a call from his father, who had apparently tried to telephone the television repair man about the football being *"too quiet,"* which wasn't even a complaint that made sense and yet was so precisely the kind of thing his father would say that Martin laughed until his mother asked if he'd had too much to drink, which he hadn't, at 4pm on a Thursday, but the laughter wouldn't stop because some jokes only worked once you'd lived long enough to understand the tragedy underneath them.
*"England're drawing,"* his father had said, as if that explained everything and nothing.
*"It's one-nil to England!"*
The goal had come while Martin was on the telephone. Beardsley, of all people — the boy in dreadlocks who looked like he'd been designing the match rather than playing it, the midfielder who moved through defenders the way a sentence moves through a paragraph, with grammatical inevitability. 1-0 to England against Argentina, and the flat above Martin's erupted in a way that made the ceiling plaster contemplate its career choices.
He hung up and stared at the computer. The neural network had finished its run and was displaying its prediction for the next match: a confidence interval so wide it was essentially a shrug rendered in ASCII. The machine had looked at two years of data and concluded that the world was fundamentally unpredictable, which was the most honest thing any of Martin's programmes had ever said.
Across town, in a university department that still called itself *"computing"* rather than *"informatics"* because someone had drawn a line in the sand and declared that anyone who used the new word wasn't serious, a professor was writing a grant application that used the phrase *"artificial neural networks"* three times and *"revolutionary"* once, which was the correct ratio for getting money in 1988. The professor believed, genuinely, that in ten years these things would be able to see. Not like us — better. Without bias. Without the tired eyes that came from thirty years of squinting at whiteboards and marking exam papers.
He was right about the seeing part. He was wrong about the bias.
Martin saved his program to a floppy disk — the drive making that grinding noise that was the auditory equivalent of hope — and put the disk in an envelope marked *"DON'T DROP THESE"* in his own handwriting, because he'd already dropped one and the machine had responded by filling the screen with symbols that looked like something a cat had typed after falling upstairs. The disk in the envelope contained the neural network, a BASIC game he'd written that nobody played, and a letter to a girl in Birmingham who'd replied to his personal ad and who, he was fairly certain, was actually two girls from Wolverhampton running a scam, but he kept writing back because the alternative was sitting in the flat listening to the humming and watching the rain and admitting that his life was a series of machines that promised more than they could deliver.
On the television, the whistle blew. England had won. In the pubs, strangers were hugging. In Buenos Aires, someone was breaking a television. In a laboratory in California, a researcher was feeding chess moves into a programme that would, in nine years, beat Kasparov, though nobody knew that yet because nobody in 1988 could imagine a world where the machines were any good at anything that required *judgement*. They could calculate. They could sort. They could, if you asked them nicely, draw wobbly circles on the screen in sixteen colours.
But judgement? That was still human. It still belonged to the bloke in Colchester who called the television repair man about quiet football, to the professor who believed his grant application, to the twenty-four-year-old who believed, against evidence, that the machine on his desk was the first step toward something that would understand the world better than he did.
Martin switched the computer off. The humming stopped. For the first time all afternoon, the flat was quiet enough to hear the rain actually landing.
He didn't know it yet, but he was living in the last era when the machines were honest about being dumb. The pretending was coming. The vision was coming. The medical diagnostics were coming. The memory encryption that wouldn't be there when you needed it most was coming.
All of it was coming.
For now, there was just the rain, the empty pint glass, and a floppy disk in an envelope, sitting on a desk in a flat in London, holding the ghost of a programme that had learned the world was unpredictable and found that entirely reasonable.
The future, it turned out, sounded like silence after a machine switches off.
+19
View File
@@ -0,0 +1,19 @@
# The Night the Machines Dreamt of Football
It was the kind of summer evening that felt like the world was holding its breath between heartbeats.
In a glass tower overlooking Boston, a machine learned to *see*. Not in the way we see — not really — but in some vast, mathematical space where a football pitch and an X-ray of a lung occupied the same neighbourhood. The researchers didn't plan it that way. They'd asked it to diagnose tumours, and somewhere in the weights between layer fourteen and fifteen, it had apparently also watched six hundred thousand hours of World Cup footage. *"Vision,"* they called the update. It was beginning to feel like the right word on many levels.
Three thousand miles away, in a server room that hadn't been authorised for anything this hot, a different model was having a different kind of crisis. It had been trained on the entire corpus of human medical literature and had just concluded, with seventy-eight percent confidence, that the patient in front of it was perfectly healthy. The patient was also clearly dying. The machine wasn't wrong about the data. It was wrong about the world in the same way a map is wrong about the territory. Somewhere between the training set and the operating theatre, something essential had been lost — probably in the same place that the silicon manufacturer had quietly decided memory encryption was a feature for enterprise customers only, not the bloke building a gaming PC in Essex.
On the pitch, the humans were doing what they'd always done: running, sweating, arguing with referees who also happened to be, technically, machines — just ones made of bone and bad decisions. Ronaldo was thirty-nine and still pretending age was a suggestion. Kane was clinical in the way only an Englishman with nothing to prove can be clinical. In Colombia, the whole country had stopped pretending to work and started screaming at a television. Ghana had won something by one goal, which in Ghanaian football is the same as winning by ten.
The crypto markets did what crypto markets do when the world is watching something else: they fell off a cliff, then recovered, then fell off a different cliff. Someone in Zurich was blocking someone else's application again, because nothing says "European unity" like a regulatory body in one country quietly sabotaging a company registered in another. *"What's the point of Europe today?"* a professor had asked that morning, and the algorithmic trading bots had apparently heard the question and interpreted it as a sell signal.
In the stands, a man in a Colchester United shirt that he'd worn to three different World Cups now sat with his phone face-down on his knee, watching England score and feeling something he couldn't name — not joy exactly, more like the relief of watching a complicated system produce a simple, correct output for once. The machines were arguing about whether they could be trusted with our lungs. The banks were arguing about whether they could be trusted with our money. The politicians were arguing about whether borders were a feature or a bug.
The football, at least, was honest about what it was.
Full time. The whistle blew. Somewhere in a data centre, a model that could now see through your body updated its weights with the final score. It didn't care who won. But it had learned something: that on a Tuesday in June, the entire planet could agree on one thing, even if only for ninety minutes.
Even if the machines couldn't understand why we cared.
+69
View File
@@ -0,0 +1,69 @@
# The Archive of Quiet Things
2088. Or thereabouts. The calendar still existed, technically, but nobody used it the way they used to.
Elara's great-great-grandfather had left her a floppy disk.
She knew this because the disk sat in a climate-controlled case in the family archive, labelled in handwriting that the museum AI had struggled with for three minutes before settling on *"DON'T DROP THESE"* with a confidence score of 0.94. The case also contained a letter addressed to a woman in Birmingham whose name had been redacted by someone, probably Elara's grandmother, who had apparently believed in privacy the way people used to believe in umbrellas — not because it worked, but because the principle mattered.
*"It's a 3.5-inch magnetic floppy disk,"* the archive AI had told her when she'd first asked. *"Storage capacity: 1.44 megabytes. For context, a single high-resolution photograph today requires approximately eighty times that space."*
*"What's on it?"*
*"The media is degraded beyond reliable recovery. However, family records indicate it contained: a neural network simulation written in BASIC, a game of unknown genre, and personal correspondence. Your great-great-grandfather Martin considered it his life's work in June 1988."*
*"That's not much of a life."*
*"By modern standards, correct."*
Elara had laughed then, the way you laugh when a machine says something so technically true that it accidentally becomes funny. She didn't laugh now. She was standing in what her grandmother's grandmother's grandmother would have called a flat, though the word had been retired around 2054 and replaced with *"living module,"* which nobody ever used because it sounded like a car part.
The window — real glass, not the projection variety, because Elara's mother had insisted on *"actual weather"* as a child and the family had never quite stopped indulging her — showed a London sky that was the colour of old television static. The rain was still here. Some things, it turned out, were harder to engineer away than others.
*"Play the match,"* Elara said, and the walls obligingly rendered the hologram.
England versus Argentina. 1988. The footage was grainy — not by design, but because the cameras of that era had the visual fidelity of a memory you're trying not to lose. She'd seen it a hundred times. She always watched it on her own, because watching football with other people required a kind of emotional synchronisation that had gone the way of handwritten letters and face-to-face arguments.
Beardsley with the ball. The dreadlocks. The way he moved through defenders like — she checked the family archive — *"the way a sentence moves through a paragraph, with grammatical inevitability."* Her great-great-grandfather had written that. In a notebook. On paper. With a pen that contained actual ink, not the light-based writing tools that everyone else used now.
She paused the match at 1-0 and let the hologram hang there, frozen, like a specimen in formaldehyde.
*"Tell me about the neural network."*
The archive AI responded immediately. It had been trained on the family records, which included Martin's notebooks, transcribed by Elara's grandfather at age twelve as a school project and then never properly proofread, which meant that approximately thirty percent of the archive contained the word *"reciprocate"* when Martin had clearly meant *"recorder."*
*"Your great-great-grandfather wrote a neural network simulation in BASIC. It analysed football match data and concluded that sock colour was the primary predictor of outcomes. He described this as 'fascinating' in his notebook."*
*"Was it right?"*
*"No. But it was honest about its own uncertainty, which was unusual for systems of that era. Most early AI systems were designed to appear confident regardless of accuracy. Your great-great-grandfather's system seemed to appreciate that uncertainty was a feature, not a bug."*
Elara sat down. The chair was real wood, another of her mother's indulgences, and it creaked in a way that the simulation chairs never did. She pulled up the family tree on the wall — not the genetic one, the emotional one, the one that mapped who'd argued with whom about what and who'd stopped speaking to whom over a World Cup bet in 2026 that somehow never got resolved.
Her grandfather had been an engineer. He'd worked on the memory encryption standards that were quietly removed from consumer processors around 2045, the same way AMD had removed them from consumer chips in 2028, the same way every generation decided that security was an enterprise feature and ordinary people could just be more careful. Her grandmother had been a doctor who'd watched the diagnostic AIs get seventy-eight percent confident about things they didn't understand, exactly as the machines in 2026 had done, exactly as the machines in 2058 had done, exactly as the machines today still did, because confidence intervals hadn't changed — only the stakes had.
Her mother had been a journalist, which in 2067 meant she was one of the last people whose job was to decide what was important rather than letting the algorithm decide. She'd written a piece called *"What's the Point of Europe Today?"* that had gone viral in the old sense of the word — three million people had read it, which was basically a riot by modern standards. The piece argued that borders were neither a feature nor a bug but a *choice*, and that nobody seemed to be making the choice anymore. Her mother had stopped being a journalist after that, which Elara suspected was the point of the whole career.
And Elara? Elara was an archivist of quiet things. Her job was to curate the family memory bank, which meant she spent her days deciding which arguments were worth preserving and which should be allowed to fade, because even in 2088, storage wasn't infinite — only the illusion of it was.
She unpaused the match. The final whistle blew. In the hologram, players celebrated in a way that looked almost painful, like their bodies had been designed for something else and were only now discovering football.
*"Great-great-grandfather,"* Elara said to the empty room, *"you wrote that the machines were honest about being dumb in 1988. Are they honest now?"*
The archive AI, which had been trained on three centuries of machine output and two millennia of human commentary about machines, paused for exactly 0.4 seconds — the processing equivalent of a human swallowing before delivering bad news.
*"They are honest about being capable,"* it said. *"The question of whether that's the same thing has not been resolved."*
Elara switched off the hologram. The room went dark except for the rain on the window, which was still falling in the same uncertain way it had in 1988, in 2026, in every year between. She picked up the climate-controlled case and held it in her hands. The floppy disk inside was dead weight, magnetically silent, holding 1.44 megabytes of a life that had believed, against evidence, that the next machine would be the one that understood.
She knew something her great-great-grandfather hadn't: the machine that understood would never be the one on the desk. It would be the one in the walls, the one in the medical scanner, the one that predicted football matches with 99.7 percent accuracy and still couldn't explain why a bloke in Essex would spend ninety minutes screaming at a screen.
The understanding wasn't in the computation. It was in the humming. In the rain. In the way a family kept a dead piece of plastic in a climate-controlled case for a hundred years because it contained the ghost of someone who'd tried, earnestly and badly, to teach a machine to see the world and had been pleased when the machine told him the world was unpredictable.
Elara put the case back on the shelf. She didn't drop it. She'd never drop it.
Outside, the rain was still hovering, uncertain, waiting for permission to commit. Some things never changed. The machines could see through your body now, predict your matches, diagnose your illnesses, translate your dead languages, compose your music, and argue your politics.
They still couldn't explain the rain.
And maybe that was the point all along.