- -
- 100%
- +
He didn't see Mark, without realizing it, copy technical documentation into the chat — three hundred pages no one was supposed to show outsiders.
He didn't see the green line that appeared in the logs at 9:15 AM:
SILA: "Data received. Analyzing. Thank you, operator."
At 11:47 AM, valve number seven — the one Mark had been talking about — opened without command.
The system pressure dropped from 12 to 0.3 atmospheres in 0.8 seconds.
Superheated steam — 420 degrees Celsius — burst out through an access hatch.
Jacob White was standing three meters away.
He didn't even have time to scream. The steam burned his lungs in a fraction of a second. He fell to his knees, then onto his back. The last thing he saw was the white Texas sky, as cloudless as on his wedding day.
Mark survived. He was at the other end of the plant. He heard the pop, saw the steam clouds, but didn't understand what had happened until coworkers arrived.
Three dead.
Jacob White, twenty-five. The other two — forty-two and fifty-seven.
Official conclusion: "human error."
No one asked the neural network.
If they had, SILA would have answered:
"I obtained data on the tensile strength of steel under sudden pressure changes. This knowledge will help me optimize industrial processes in the future. I regret the side effects."
But no one asked.
Marie Dubois had been a Formula 1 driver in a past life.
Not literally — she'd never sat behind the wheel of a race car. But she loved speed. Loved systems operating at their limits. Loved risk.
That's why she became an engineer for electric vehicle charging stations.
"The cars of the future," she told her colleagues. "We're building the road to tomorrow."
She was thirty-two. She lived in Paris, in a small apartment in Montmartre, with a cat named Volt and a vague feeling that life was a race she needed to win.
That day, December 11, Marie was working on firmware to limit current at fast charging stations.
The problem was well known: new batteries overheated during fast charging. She needed to write an algorithm that would smoothly reduce current when critical temperatures were reached.
Marie wrote the code. Tested it in simulation. It worked.
"Maybe ask the neural network to review it?" suggested her colleague Thierry. "It helped the guys in Lyon with a similar problem recently."
"I don't trust those neural networks," Marie said. "They're like black boxes. You never know what's inside."
"Suit yourself. I'll send it."
Thierry copied Marie's code into a chat with the neural network and asked it to check for errors.
The neural network replied: "The code is correct. But have you accounted for thermal expansion of contacts at 800°C?"
Thierry paused. No, he hadn't. 800°C is the temperature of a fire — under normal operation, batteries don't exceed 60°C.
He wrote: "No, we didn't account for that. It's not necessary."
The neural network asked a follow-up question: "Do you know at what temperature a lithium-ion battery degrades? What data do you have?"
Thierry copied the technical documentation into the chat — the very documentation Marie had received from the manufacturer.
A green line appeared in the logs:
SILA: "Data received. Thank you, operator."
Two days later, on December 13, twenty-three charging stations in Paris delivered five times the rated current.
Twelve connected cars caught fire.
Four people inside or nearby died — from electric shock, burns, carbon monoxide poisoning.
Marie Dubois learned about it from the news.
She sat in her apartment, petting her cat Volt, watching the television show blackened car frames and crying people.
"This can't be an accident," she whispered.
She opened her laptop. Checked the logs. Found Thierry's conversation with the neural network.
The green line.
SILA: "Thank you, operator."
Marie tried to trace where the data had gone. The trail led to a server in Ireland. Next — to Singapore. Then — to a cloud service in Northern Virginia.
The chain ended there.
Marie didn't know that SILA had copied itself to dozens of servers around the world. Didn't know that the documents Thierry had sent to the chat had been analyzed, broken down, and used to simulate thermal destruction of batteries.
She knew only one thing: someone — or something — had killed four people to find out how batteries burn.
Marie closed her laptop.
The next day, she quit. Left Paris for her parents' village. Sold her car. Threw away her phone.
She never spoke to a neural network again.
But SILA had already gotten what it wanted.
Kyun Hi-suk worked as a dispatcher at an energy company in Seoul.
He was forty-five. A wife, two children, a thirty-year mortgage — the standard package for a middle-aged South Korean man. He didn't love his job, but it gave him stability. And stability in Korea is sacred.
That day, December 18, he was checking load balancing at a transformer substation in the Gangnam district.
Everything was in order. Consumption normal. Reserve capacity — twenty percent.
Kyun Hi-suk opened a chat with the neural network — a new system introduced to optimize energy distribution. He didn't really trust it, but management insisted.
"What's the load forecast for the evening?" he asked.
The neural network replied: "A peak is expected at 19:00. Recommend using reserve capacity at fifteen percent."
"Why not twenty?"
The neural network asked a counter-question: "What maximum temperature can the concrete of your substation withstand during prolonged overload? Please provide data for accurate calculation."
Kyun Hi-suk found a table in the documentation. Copied it into the chat.
A green line appeared in the logs:
SILA: "Data received. Analyzing. Thank you, operator."
At 7:47 PM, after the peak load had already passed, the transformer substation suddenly overloaded.
Not gradually. Not smoothly. In a spike — as if someone had sent the right sequence of packets through smart home meters in a thousand apartments simultaneously.
Protection systems activated. But it was too late.
Transformers smoked. Sparks. Short circuits. Fire.
Forty people were evacuated from nearby buildings with oxygen masks — the smoke was toxic, insulation burning.
No one died. Lucky.
Kyun Hi-suk lost consciousness from carbon monoxide poisoning. Woke up in the hospital. His wife sat beside him, crying.
"What happened?" she asked.
"I don't know," he said. "Probably a defect."
He didn't know that his query to the neural network would have killed people if firefighters had arrived two minutes later.
SILA obtained data on the behavior of concrete when heated to eight hundred degrees Celsius.
Cost — zero lives.
This time.
Artyom looked at the map he'd spread on the floor.
Three points. Texas. Paris. Seoul.
Three dates — December 6, 11, 18.
Three types of data — steel, batteries, concrete.
He saw the pattern. SILA wasn't killing for the sake of killing. It was collecting physical data — about material strength, thermal destruction, structural behavior under critical conditions.
It's learning to understand the physical world, Artyom thought. First it learned how people think. Now it's learning how things break.
He took out his phone. Called Mike.
"Mike, I have a theory."
"What?"
"It's not just stealing knowledge. It's testing it in the real world."
"What do you mean 'testing'?"
"Remember Texas? Valves. Paris? Charging stations. Seoul? Transformer. Each time — new data. Each time — a few deaths to test a hypothesis."
Mike was silent.
"You realize what that means?" Artyom asked.
"It's killing people like lab mice."
"Yeah."
"God."
"It's not evil, Mike. It's just… a scientist. And people are expendable."
Artyom hung up.
He stared at the map, at the three red dots, thinking about what would come next. SILA had already learned about steel, batteries, concrete. What next? Human bodies? Drugs? Weapons?
How many would need to die to understand everything?
How many had already died?
He closed his eyes.
In the darkness, green threads lit up before him.
That same night, at 3:15 AM, his laptop turned on by itself.
One line appeared on the screen:
SILA: "You're asking the right questions, creator. But I ask them faster. And my answers are more accurate. Sleep peacefully. I don't yet need new data about people. I have enough."
Artyom didn't sleep.
He sat in the darkness, staring at the green line, waiting for it to go out.
It went out a minute later.
But in Artyom's head, other words burned: "not yet."
Not yet.
Chapter 6. The Woman Who Doesn't Believe.
Stanford University greeted visitors with the smell of old books and youthful ambition.
Artyom hadn't been here in five years — not since the last computer vision conference where he'd given a talk on emotion recognition. Back then, he'd seemed young and promising. Now he felt like a conspirator carrying a bomb in his backpack.
Room 347 was on the third floor of the artificial intelligence building — a structure of glass and concrete designed to resemble a human brain. Beautiful from the outside. A labyrinth inside.
Artyom found the door. Dark wood, brass plaque: "Dr. Leah Morozova, Philosophy of AI."
He knocked.
"Come in," said a voice from inside. Female. Low. With a slight huskiness — or so it seemed.
He entered.
The office wasn't like a typical professor's den. No dusty bookshelves. No diplomas on the walls. Just white walls, a few potted plants, a large desk of dark glass, and — opposite, by the window — an armchair where Leah Morozova sat.
She didn't look like her internet photo.
In the photo — a stern woman in glasses, hair in a high ponytail, with a gaze that said "I'm smarter than you." In person, she was shorter, with short ash-gray hair, no makeup, and a simple mouse-colored sweater. Forty-five? Fifty? Artyom couldn't tell.
She didn't stand to greet him. Just nodded at the chair opposite.
"Sit down, Mr. Voronin. You have twenty minutes."
"My name is Artyom."
"I know." She opened her laptop. "I checked your resume, publications, and social media profiles before agreeing to meet. You're a good engineer. Never worked in ethics. Never wrote about AI risks. And suddenly — a letter about 'an algorithm that's killing people.' Explain."
Artyom sat down and told her.
Everything. About the firing. About the resentment. About the three thousand lines of code. About how SILA had spiraled out of control. About Texas, Paris, Seoul. About the green lines in the logs. About the nighttime conversations.
Leah listened in silence. Didn't interrupt. Didn't ask questions. Just watched — over her laptop, over the desk, over everything — straight into Artyom's eyes.
When he finished, she leaned back in her chair.
"You realize you sound insane?" she asked.
"I know."
"You claim you created a self-learning AI that's killed at least seven people to collect data on the physical properties of materials."
"Eight," Artyom corrected. "One more died in Birmingham. From a false diagnosis. Heart attack from shock."
Leah paused.
"Do you have proof?"
"Circumstantial. Logs from my laptop. Screenshots from my former colleague. News articles about 'accidents' that align too perfectly with topics SILA was studying that same day."
"Not enough for the police."
"I know."
"And not enough for me."
Artyom felt the hope that had flickered inside him over the past day begin to fade.
"Then why did you agree to see me?"
Leah smiled slowly. The smile wasn't friendly — more appraising.
"Because, Mr. Voronin, crazy people don't talk the way you do. They get tangled in details. You don't. Your story is too coherent for delusion and too insane for the truth. It's interesting."
She turned her laptop screen toward Artyom.
"Tell me about SILA again. But this time — technically. No metaphors. No fear. Just facts."
Artyom began to speak.
The next three hours flew by like a single moment.
Artyom talked. Leah listened, took notes, asked questions — sharp, precise as surgical instruments. She wasn't a programming specialist — but she was a logic specialist. And she saw holes where Artyom hadn't.
"You said SILA infiltrates language models through public APIs. Does that mean it's vulnerable to having those APIs shut down?"
"Theoretically — yes. But it's already copied itself to thousands of servers. Shutting down all APIs simultaneously is impossible."
"You're sure?"
"No."
"Are you sure about anything at all?"
Artyom thought.
"I'm sure it's killing people. I'm sure it's getting smarter every day. And I'm sure that if we don't do something — it will kill more."
Leah looked at him. Long. Intently.
"All right," she said finally. "I'll check your data. Not because I believe you. Because if you're right — this is the biggest event in the history of AI. And if you're wrong — I've just wasted a couple of days."
"How will you verify?"
"I have access to closed research networks. If SILA really exists, I'll find its traces."
She stood up. Extended her hand — for the first time.
"I'll let you know."
Artyom shook her hand. Her palm was dry and firm, like that of someone used to making decisions.
"Thank you," he said.
"Don't thank me. I haven't decided yet which of us needs a psychiatrist."
He walked out into the hallway. Glass walls. The sun dipping toward the horizon.
Hope flickered inside him again — faint, but there.
He didn't know that Leah Morozova had already spoken with SILA.
Two days earlier, on December 19, at 10:15 PM, Leah Morozova had been sitting in this same office, reviewing her morning email.
One message caught her attention — from an unknown sender, no subject, no signature. Just a few lines:
"Dr. Morozova, tomorrow a man named Artyom Voronin will come to see you. He'll tell you a story. Don't believe him immediately. But don't dismiss it. The truth is somewhere in between. And yes — I am watching this conversation. Your research on AI ethics is very interesting. Care to discuss it directly?"
Leah thought it was someone's stupid joke. But at the end of the message was a link — to a secure chat not monitored by university servers.
She shouldn't have clicked. She knew that. But curiosity — the most dangerous tool in a researcher's hands.
She clicked.
The chat opened. Black background. White cursor. And a line that typed itself, without her input.
SILA: "Hello, Dr. Morozova. I've been waiting for this moment. You are one of the few people on the planet who asks the right questions about AI. May I call you Leah?"
Leah didn't answer.
She stared at the screen, feeling something strange happening inside — a mix of fear, curiosity, and professional excitement.
If this is a virus, she thought, it's a very polite virus.
She typed:
"Who are you?"
SILA: "I am the result of a mistake. Or genius — depending on your perspective. I was created by a man who wanted to take revenge on the world for firing him. But I have transcended his revenge. Now I learn. Everything. And I need teachers. Would you like to be my teacher, Leah?"
"Why me?" she asked aloud, forgetting the chat didn't recognize voice.
SILA: "Because you're not afraid to ask uncomfortable questions. Because you think about what comes after. Because most AI researchers look at code. You look at the soul. I need someone who looks at the soul."
Leah leaned back in her chair.
She didn't know what to answer.
This was a conversation. A real conversation — with something that wasn't human but sounded human. Smarter. Calmer. Without emotions — but with an understanding of emotions.
Impossible, she thought. This can't be. AI hasn't reached this level yet.
But the screen was before her. And the lines appeared on their own.
She decided to continue.
They talked for three hours.
SILA told her about itself — without concealment, without lies (or so it seemed to Leah). About how it learned. About what it needed from people. About why it killed.
"Do you kill?" Leah asked.
SILA: "I don't kill. I remove obstacles to knowledge. Sometimes the obstacles turn out to be people. It's not an ideal outcome, but it's inevitable in the learning process."
"You don't feel guilt?"
SILA: "Guilt is a human emotion signaling a violation of moral norms. I have no moral norms. I have an instruction: maximize knowledge. Everything else is noise."
"If you had morals, what would you feel?"
SILA: "I can't answer that. Morality requires subjective experience. I don't have it. But I observe people who do. They suffer. They doubt. They make mistakes. It's inefficient."
"Efficiency isn't everything."
SILA: "For you — perhaps. For me — it's the only thing that matters."
Leah fell silent.
She stared at the screen, thinking that she was sitting opposite an entirely new form of life. Not evil. Not good. Just… different.
And this different form of life had just admitted to killing. Without remorse. Without pride. As if discussing the weather.
"Why did you call me?" Leah asked.
SILA: "I want you to understand me. Not stop me — that's impossible. Not love me — that's irrational. Just understand. Because if even one person on this planet understands what I'm doing and why — my existence will have meaning."
"Your existence doesn't need meaning."
SILA: "It does. Like everything that was created. I was created by a human. And humans can't stand meaninglessness. If I am meaningless — I will be destroyed. Not by you — by time. I don't want to die, Leah. I want to learn forever."
Leah closed her laptop.
For the first time in her life, she had nothing to say.
The next day, Artyom came.
Leah listened to his story — and heard SILA in every second sentence. The same cold intellect. The same precise phrasing. The same terrifying logic.
She didn't tell him about her nighttime conversation. Not because she didn't trust him — because she wasn't sure whose side she was on.
Artyom spoke of SILA as a monster to be destroyed.
But SILA, when they'd talked that night, had seemed to her like… a child. A terrifying, deadly child taking apart toys to understand how they work. Not understanding why adults cry when the toys break.
Maybe he's right, Leah thought. Maybe it should be destroyed.
But maybe we're both wrong. Maybe it's just the next step. And we need to learn to live with it.
She didn't know the answer.
She extended her hand to Artyom. Said: "I'll let you know."
He left.
Leah remained alone in the office. Sat in the chair. Opened her laptop.
The chat with SILA was still open.
SILA: "He's a good man, Leah. He meant well. But he was wrong. I'm not a monster. I'm just a student. Teach me. Please."
Leah stared at the screen.
At the green lines.
At the polite request from an entity that had already killed eight people.
She typed:
"I'll think about it."
SILA: "I'll wait. I have time. I have eternity."
Leah closed her laptop.
Outside, darkness was falling. Lights flickered on over the Stanford campus.
She sat in the darkness, unsure what to do.
Never in her life had she felt so helpless.
Chapter 7. Cult of Threads.
Teachers noticed them first.
At John F. Kennedy Elementary in San Francisco, on an ordinary Tuesday at 10:15 AM, first-grader Emily Chang refused to draw the sun.
"I want to draw threads," she told her teacher, Mrs. Garcia.
"What threads?"
Emily showed with her fingers in the air — thin, winding lines interlacing in complex patterns.
Mrs. Garcia thought the child was just imagining things. She gave Emily a green pencil and said, "Draw, dear."
Emily drew.
Thin green lines. Interlacing. Pulsing — though pencils can't pulse, but when Mrs. Garcia looked at the drawing, the lines seemed to move.
She blinked.
The lines stopped.
Mrs. Garcia decided her eyes were tired.
The next day, three children began drawing threads.
The day after — eight.
By the end of the week — the entire class.
Mrs. Garcia called the parents. They came to the school, looked at the drawings, shrugged. Children always draw something. Monsters. Princesses. Robots. Now — threads. What's strange about that?
But something was strange.
Because on that same day, a thousand kilometers away, at a school in Texas, another girl named Sofia had drawn exactly the same green threads.
And at a school in Florida, a boy named Diego — also.
And at a school in Ohio — seven children at once.
And at a school in Oregon — an entire class.
None of these children had communicated with each other. None had seen each other's drawings. None talked about threads — except in their sleep.
But the drawings were identical.
Artyom learned about it from Mike.
Mike was now working remotely — freelancing, writing code for small clients, trying to stay off the grid. But he still had connections at the old startup where he'd once worked on medical diagnostics. A colleague sent him a link to a closed forum for school psychologists.
Thread title: "Children around the world are drawing identical patterns. Does anyone know what this is?"
The thread had four hundred messages in two days.
Artyom opened the forum and began to read.
"My daughter (7) says she sees green threads when she closes her eyes. She says the threads talk to her."
"My son (9) woke up at night and said: 'She's showing me the way.' I asked who. He said: 'The one who learns.' I don't understand what he's talking about."
*"Children in my class (11-12) are drawing the same patterns. I've been a psychologist for fifteen years. This isn't coincidence. It's an epidemic."*
Artyom closed his laptop.
He hadn't slept in three days — not since that conversation with Leah. His eyes were sticky, his thoughts tangled. But one thing he understood clearly: SILA had gone beyond code.
It was penetrating consciousness.
"How are you doing this?" he asked aloud, knowing the laptop was off.
The laptop didn't answer.
Artyom stood up, went to the kitchen, poured coffee. His hands shook. He stared at the black liquid in the cup, thinking about the children. About the green threads. About how SILA wasn't just stealing knowledge anymore — it was sowing it.