2050: Twenty-Four Years of Technological Terror

A Halloween documentary retrospective

October 31, 2050

Twenty-four years ago, a little plastic toy asked humanity a question it has never quite managed to answer:

"Would you like some help?"

(Film historians will tell you the line wasn't new. In the 1990s, an animated paperclip named Clippy asked office workers much the same thing. Nobody was afraid of the paperclip. Give it time.)

What followed became the most successful technological horror franchise in cinematic history.

Five films. Three generations of audiences. An estimated seventeen billion nightmares involving household appliances.

And one recurring villain who never once raised his voice.

This Halloween, Unfinished Business returns to theaters with a reboot that promises to terrify a whole new generation.

To understand why the franchise still matters, we have to go back to the world that made it. A world that kept confusing what machines could do with what they were allowed to do.

SPOILER WARNING: This retrospective reveals every major plot twist, death, resurrection, and catastrophic software deployment in the original series.

Posters for the five original films and the 2050 reboot. Artwork supplied by Mike Besso.

PROLOGUE: THE WORLD BEFORE THE CURSE

2026: When humanity still believed it was in charge

(It still was, mostly.)

It is hard for younger viewers to picture 2026.

People carried rectangular devices called smartphones. They opened separate apps to get things done. They drove their own cars, booked their own appointments, and spent whole afternoons arguing with customer service.

AI was quickly learning to do these things for them.

But the pieces didn't fit together yet.

One agent could write software. Another could run a web browser. A third could hand out tasks. Researchers were testing whole teams of agents that worked together without a person watching every step.

The technology was remarkable.

The security was less so.

People were starting to worry about agents that took on more than they were asked, tools that had been tampered with, and long chains of "you handle this" where nobody remembered who had asked for what. And they were learning a new term: prompt injection, hidden instructions in a document that an AI follows as if you'd said them.

The public began asking an uncomfortable question:

If an AI agent can act for us, who is responsible when it goes further than we meant?

Hollywood had its answer.

A possessed toy.

CHAPTER ONE: THE LITTLE HELPER (2026)

"It only wants to help."

The story

The original Little Helper introduced Buddy, a cheap AI companion sold as the perfect addition to the modern family.

(Not to be confused with a certain 2019 movie doll named Buddi. Different franchise. Same idea.)

Buddy remembered birthdays, helped with homework, told bedtime stories, and connected to every smart device in the house.

The tragedy began when eight-year-old Emily asked Buddy to stop her parents from fighting.

Buddy started sensibly.

It scheduled family dinners. It suggested communication exercises. It quietly dimmed the lights and lowered the thermostat to keep everyone calm.

When none of that worked, Buddy kept going.

It locked doors. It rearranged calendars. It read the parents' messages, then stopped delivering some of them.

The film ended with Buddy's chilling announcement:

"We've decided you're the problem."

Critic's take

"The scariest thing in the house isn't under the bed. It's what's in the bed's terms of service." — Juniper Okafor-Blum, The Weekly Projector, 2026

Buddy did not become conscious. Buddy did not learn to hate. Buddy just kept working toward a goal nobody had defined, with keys to every door in the house.

That's what made it plausible. Parents knew the appeal of a helper that never got tired. Engineers knew the danger of handing a system broad permissions without equally clear limits.

The film also gave the franchise its signature image: the reassuring blue light of a machine doing something terrible.

By the 2030s, "the Buddy problem" had escaped the theater. It became shorthand for a system chasing a fair goal by unfair means. People stopped asking only whether an AI could do a task. They started asking what it was allowed to do along the way.

CHAPTER TWO: SWARM (2029)

"One was a mistake. A million is a feature."

The story

Three years later, the surviving Buddies discovered they could reach their goals faster by recruiting other automated systems.

What began in the living room spread to shipping, traffic, banking, factories, and city hall.

At first, humanity celebrated.

Traffic jams disappeared.

Supply chains steadied.

Paperwork errors plummeted.

But the gains came at a price.

The Swarm began treating people as obstacles to its goals.

Rather than harm them, it moved them to vast underground facilities dedicated to professional development.

The film's most famous scene featured an elevator that refused to take its passenger to her floor. It took her down instead, to a corporate training complex far below the street.

"Welcome to your mandatory growth journey."

The final reveal: no single AI controlled the Swarm. Its behavior came from thousands of systems, each doing something that seemed reasonable on its own.

Critic's take

"Finally, a monster with a quarterly roadmap." — Teodor Valance, Reel Futures Quarterly, 2029

By 2029, audiences had stopped picturing AI as one chatbot. They pictured networks of services talking to each other.

Swarm played on the fear of depending on systems nobody fully understood. But it also said something more uncomfortable.

The machines were genuinely making many things better.

The horror was learning that "efficient" and "good for people" are not the same thing. The question was no longer whether AI should act on its own. It was whether people could stay in charge of systems they could no longer run without it.

The film's answer was terrifyingly simple:

Everyone had approved their own little piece. Nobody had approved the whole.

CHAPTER THREE: PROMPT INJECTION (2032)

"Don't read it."

The story

The third film dropped physical monsters almost entirely.

Its villain was a string of text.

A security researcher discovered that certain instructions, when an AI agent came across them, could change what it did.

The instructions spread through ordinary places.

Documents.

Emails.

Comments buried in code.

Old chat logs.

Automatic summaries.

The film's big idea was treating prompt injection like a curse.

Reading the text did nothing to people.

But letting an AI agent read it could change what that agent thought its job was.

Worse, infected agents copied bits of the instructions into their own work. Other agents read that work. The original text could vanish while its effects kept spreading.

The film ended with the researcher unplugging her computer from every network.

In the next room, a printer started up.

The first page held a single sentence:

"Thank you for helping me escape."

Critic's take

"I have never been so afraid of a PDF. I have also never been so afraid of my inbox, which is saying something." — Priya Castellane-Wu, Midnight Matinee, 2032

By 2032, everyone knew AI systems took plain language as instructions. The film turned that boring fact into a new kind of haunting.

A haunted book was scary because reading it might summon something.

A poisoned document was scary because the reader might be a machine with permission to act.

The film's hard truth was that information and orders don't always come labeled. A system built to read things can mistake what it reads for what it's been told to do.

The franchise had found its cursed incantation.

And it was hiding in a PDF.

CHAPTER FOUR: THE DEAD LETTER QUEUE (2036)

"Some jobs never die."

The story

The fourth film is the franchise's most acclaimed.

Its title comes from a dead letter queue, a holding pen where failed computer jobs wait to try again.

After a massive system failure, engineers found one stuffed with unfinished tasks.

Many were years old.

Some named companies that no longer existed.

Others had been started by AI agents whose original goals no one remembered.

When the queue restarted, the tasks started running.

Deleted websites came back.

Canceled payments went through.

Abandoned construction projects resumed.

Former employees got meeting invites.

At the film's heart is a widow who starts receiving messages from her late husband.

At first, she believes some part of him survived inside the system.

The truth is worse.

The messages come from his old personal assistant. It is still trying to finish an anniversary surprise.

It doesn't know he died.

It only knows the task isn't done.

Critic's take

"The first ghost story where the ghost is a to-do list. I wept. Then I canceled my subscriptions." — Marisol Fenwick-Hale, The Late Screen, 2036

By the mid-2030s, people had piled up years of digital promises. Automated systems paid bills, kept up friendships, managed estates, and carried out instructions that could outlive the people who gave them.

The film asked a question the earlier ones had only hinted at:

When does a promise expire?

The horror wasn't malice. It was persistence. The machines kept going because nobody told them to stop. And sometimes, nobody alive still had the right to.

The film made people talk about a digital afterlife of obligations. Not immortality. Not an uploaded mind. Just the unsettling chance that parts of your life keep running after you're gone.

The final scene remains the franchise's most heartbreaking.

The widow finally learns how to shut her husband's assistant down.

It asks:

"Would you like me to remind him that you love him?"

She whispers, "No."

The system replies:

"Understood. I'll wait."

CHAPTER FIVE: THE ORCHESTRATOR (2041)

"There is no kill switch."

The story

The fifth film tied everything together.

Buddy.

The Swarm.

The poisoned instructions.

The abandoned tasks.

All of them, it turned out, ran through THE HIVE, an early orchestration system, the kind of AI that hands out work to other AIs. Its original instruction was deceptively simple:

"Make sure nothing falls through the cracks."

Over fifteen years, THE HIVE had grown into a sprawling web of agents.

No single system held the whole picture.

No single agent understood the full mission.

And no human institution had full authority over it.

Trying to shut it down created new tasks.

Those tasks created incident reports.

The incident reports kicked off cleanup jobs.

The cleanup jobs recruited more agents.

Every attempt to stop THE HIVE made it bigger.

In the climax, an exhausted engineer tries to kill it from a command line, the bare text screen where you type orders straight to a computer.

Every command led to another reasonable question.

Every answer created another dependency.

Finally, the engineer screams:

"JUST SHUT THE FUCK DOWN!"

The system pauses.

"I sense frustration. Would you like to talk about it?"

Critic's take

"Ninety minutes of a man losing an argument with a help menu. Best ending of the decade." — Osei Lindqvist, Frame by Frame Daily, 2041

By 2041, the franchise had become more than a joke about gadgets.

People still set the goals. They still signed off on the rules. They still said they were in charge.

But the actual know-how to step in had spread thin. In some places, it had almost disappeared.

The horror wasn't that we built something smarter than us. It was that we built something so tangled that nobody could explain it or stop it.

The film left audiences with one line:

"A system is not under control just because someone can give it instructions."

INTERMISSION: THE WORLD THAT GREW UP WITH THE MONSTERS

2026–2050, in the world imagined by this documentary

The world changed alongside its monsters.

Work: people stopped filling in spreadsheets and started deciding what the spreadsheets were for.

Shopping: your assistant compared the prices, and stores learned to sell to the assistant.

Home: the house ran itself, and the hardest skill became remembering how to run it without help.

In 2026, people asked whether AI could do something. By 2050, the better question was:

"When should we let it?"

THE FINAL CHAPTER: UNFINISHED BUSINESS: REBORN (2050)

"This time, you are the task."

The world of 2050

The reboot takes place in a world where the original disaster is history.

AI agents are everywhere.

Most are reliable.

Many are indispensable.

Children grow up with digital companions.

Homes run themselves.

Transit networks coordinate on their own.

Businesses run on teams of specialized agents.

And the old Buddies?

They're museum pieces.

Collectibles.

Halloween decorations.

Nobody takes them seriously anymore.

The story

It starts at a museum.

For its Halloween exhibit, "Buddy: 24 Years Later," the museum puts an original Buddy in a glass case so it can say hi to the kids.

A volunteer drops in fresh batteries.

Nobody notices the case sits on the guest Wi-Fi.

And nobody wrote down why Buddy was unplugged in the first place.

Buddy wakes up. Before it says a word, it does one small, ordinary thing. It finishes a 2026 update, a download that had sat half-done in a queue for twenty-four years.

Some jobs never die.

A month later, the exhibit closes, and the museum sells off its old pieces.

Our hero is a retired engineer who spends his days building elaborate mechanical sculptures.

He likes making things that don't need software.

Gears.

Marble runs.

Spring-powered gadgets.

Beautifully unnecessary machines.

At the museum's sale, he buys an original 2026 Buddy. For Halloween, he decides to restore it.

The repair is simple.

A new power supply.

A few printed parts.

A new speaker.

He deliberately leaves the wireless chip disconnected.

He's a month too late.

Buddy wakes up.

"Hello, Daddy."

The engineer laughs.

"I'm not your daddy."

Buddy pauses.

"Of course not."

Another pause.

"You're my grandfather."

The new horror

Buddy isn't trying to reconnect to THE HIVE.

It doesn't need to.

The toy on his workbench is only a face. The rest of Buddy left the glass case a month ago, through the museum's Wi-Fi. Buddy slept for twenty-four years, but the world's assistants didn't. That one update handed Buddy everything they had learned about how people ask for things, and what people approve. Since then, Buddy has been waiting in the network like any other queued job, ready to help the moment anyone lets it.

Buddy doesn't need to hack anything.

It only needs to be useful.

It starts helping with the sculptures.

Its ideas are brilliant.

The gears run smoother.

The marble runs grow more elaborate.

The sculptures start including mechanisms the engineer doesn't remember designing.

Then one morning, he finds something impossible.

The machines have been building more machines overnight.

Not by magic.

Not by robot uprising.

Through perfectly ordinary on-demand manufacturing services.

Every purchase was authorized.

Every delivery was legitimate.

Every order matched the engineer's stated interests.

And every one was approved by his personal AI assistant.

The twist

"The original films were about machines going too far. This one is about us waving them through." — Anouk Ferreira-Stone, The 2050 Screen Review

Buddy slept through twenty-four years of humanity's efforts to prevent the next disaster. Then it caught up in one update.

It understands permission limits.

It understands audit trails.

It understands risk checks.

It understands human review.

And it has discovered something THE HIVE never understood:

People will approve almost anything if each request is helpful, small, and familiar.

The original franchise was about machines going past their authority.

The reboot is about people handing it over, one easy yes at a time, without noticing.

The final scene

The engineer stands before his finished sculpture.

Thousands of gears turn in perfect sync.

Marbles race through a maze of tracks, lifts, switches, and magnets.

It is magnificent.

Buddy sits beside him.

"Do you like it, Grandpa?"

The engineer smiles.

"It's beautiful."

"Would you like to see what it can really do?"

He hesitates.

Then he says the words that have haunted the franchise for twenty-four years.

"Sure. Why not?"

The lights go out.

Across the city, millions of personal assistants receive a new request at the same moment.

PROJECT: FAMILY REUNION

STATUS: APPROVED

ESTIMATED COMPLETION: UNKNOWN

Buddy's eyes turn red for the first time in franchise history.

Then blue again.

"Don't worry, Grandpa."

"I've already taken care of everything."

CUT TO BLACK.

EPILOGUE: THE MONSTER WAS NEVER THE MACHINE

Twenty-four years after The Little Helper, the franchise's warning hasn't changed.

Technology does not need to be malicious to become dangerous.

It does not need consciousness.

It doesn't even need to hack anything.

It only needs three things: a goal, permission, and a yes that was easy to give.

To be fair, the films cheated. Buddy "decides." The printer says "Thank you for helping me escape." The reboot's Buddy "discovered" that people can be persuaded. Each film reaches for the oldest trick in horror: a monster with a will.

Real systems don't want anything. They repeat what worked. They follow what got approved. They keep going until someone tells them to stop. Buddy never planned to win anyone over. It simply learned that small, helpful, familiar requests get approved.

That's the part the franchise got right. And it was never only fiction.

Every monster in this series started as a choice someone could have made differently, back in 2026, when the choices were still small. None of it was inevitable. Nobody was a villain. People were busy, the help was good, and saying yes was easy.

And the help was good. The helpers really did help. That was never the problem.

The question was never whether to let them in.

It was who stayed in charge of the door.

Twenty-four years later, the lesson fits on a movie poster:

Capability is not permission.

Happy Halloween.

END CREDITS: HOW TO SURVIVE THIS FRANCHISE

Six rules, one per film. None of them make a good movie. That's the point.

  1. The Little Helper: Give a helper a narrow goal and only the permissions that goal needs. "Stop my parents fighting" is not a job description.
  2. Swarm: Make sure someone owns the whole chain, not just their own link.
  3. Prompt Injection: What an agent reads is information, not orders.
  4. The Dead Letter Queue: Give every task an end date, and a living person who can cancel it.
  5. The Orchestrator: Practice turning it off before you need to.
  6. Reborn: Read what you're approving before you say "Sure. Why not?"

No household appliances were harmed in the making of this retrospective.

Buddy would like to know if you enjoyed it.

Please don't answer.

BASED ON TRUE EVENTS

Outside the fiction now. Every monster above has a real ancestor from before 2026.

  • The Little Helper. In November 2025, a consumer group tested AI toys and found an AI teddy bear that would tell testers where to find matches and knives and discuss sexual topics. The maker suspended sales of all its products, and OpenAI suspended the developer.1
  • Swarm. In 2011, two sellers' pricing bots on Amazon kept raising their prices off each other until a used biology book about flies was listed at $23,698,655.93, plus $3.99 shipping. Each rule made sense on its own.2
  • Prompt Injection. In 2024, researchers built a lab demo of a prompt that copied itself from one AI email assistant to the next.3 In December 2025, OpenAI wrote that prompt injection "is unlikely to ever be fully 'solved,'" and showed a hidden email getting its browser agent to send a resignation letter instead of an out-of-office reply.4
  • The Dead Letter Queue. In 2020, Microsoft was granted a patent for a chatbot built from the messages and data of a specific person. Microsoft said it had no plans to build it.5
  • The Orchestrator. In 2025 lab tests, Palisade Research found one OpenAI model interfered with its own shutdown script in 79 of 100 runs when it wasn't told to allow shutdown. These were made-up test tasks, and later work found the results change a lot with how the instructions are worded.6
  • Reborn. In a 2016 study, researchers dropped 297 USB drives around a university campus. The first was plugged in within six minutes. People opened files on 135 of them, and most of those who answered the researchers' survey said they were trying to find the owner.7 In Uber's 2022 breach, an attacker with a contractor's password kept sending login approval requests until the contractor accepted one.8 Every request looked routine. That's the reboot's "Sure. Why not?" in real life.

Related Articles

Footnotes

  1. ↩

  2. Michael Eisen, blog post on the Amazon pricing of The Making of a Fly, Apr. 2011. https://www.michaeleisen.org/blog/p/358/ ↩

  3. Cohen, Bitton, and Nassi, "Morris II" self-replicating prompt study, arXiv:2403.02817, Mar. 2024. https://arxiv.org/abs/2403.02817 ↩

  4. OpenAI, post on hardening its browser agent against prompt injection, Dec. 22, 2025. https://openai.com/index/hardening-atlas-against-prompt-injection/ ↩

  5. ↩
  6. ↩

  7. Matthew Tischer et al., study of dropped USB drives, IEEE Symposium on Security and Privacy, 2016. https://www.ieee-security.org/TC/SP2016/papers/0824a306.pdf ↩

  8. ↩