What if Rome had invented the Internet?**
Not in 1990.
Not in 1900.
In 100 AD.
Imagine Roman engineers discovering a primitive communication network capable of transmitting information across the empire within hours.
At first, it sounds like a technological upgrade.
Faster military orders.
Faster trade information.
Faster political communication.
Nothing revolutionary.
But then something happens.
Information spreads.
Institutions adapt.
Power structures change.
Science accelerates.
Religions evolve.
Wars change.
Economies transform.
And eventually...
the world of 1100 AD no longer resembles our history.
This is the question behind my latest experiment:
The Butterfly Effect
An AI-powered interactive laboratory for exploring alternative histories and causal consequences.
🔗 Open source:
https://github.com/modarresi1913/the-butterfly-effect
🦋 History Is Not a Line
We usually visualize history like this:
100 → 200 → 300 → 400 → 500 → 600 → 700 → 800
But complex systems don't really behave like that.
A tiny intervention can create a chain reaction:
Historical Event
│
▼
Small Intervention
│
┌─────────┴─────────┐
▼ ▼
Direct Effect Unexpected Effect
│ │
└─────────┬─────────┘
▼
Social Change
│
▼
Political Change
│
▼
Technology Shift
│
▼
New Civilization
The further we move through time, the further the new timeline can diverge.
That's the Butterfly Effect applied to history.
🌐 The Experiment: Rome Gets the Internet
Let's run a thought experiment.
Initial Conditions
Year: 100 AD
Location: Roman Empire
Intervention: A primitive global communication network
Access: Public
Simulation: 1,000 years
We press:
SIMULATE
And the timeline begins to diverge.
🏛️ Year 100 — The Roman Network
The Roman Empire already has something extraordinary:
- roads
- postal systems
- administrative bureaucracy
- military logistics
- engineering expertise
Now add a high-speed information network.
The first-order consequences are obvious.
Military commands travel faster.
Trade information becomes available almost instantly.
Provincial governors communicate with Rome more efficiently.
But this is only Layer One.
The real butterfly effect hasn't started yet.
🧠 Year 200 — Information Becomes Power
Merchants discover that the network is useful.
Then scholars.
Then politicians.
Then ordinary citizens.
Information begins moving faster than people.
Libraries start maintaining shared knowledge repositories.
Researchers in Alexandria can exchange ideas with scholars in Rome.
Engineers in Hispania can communicate with architects in Greece.
The empire has accidentally created something much more important than a communication system.
It has created a civilizational nervous system.
⚔️ Year 300 — The First Information War
Now politics adapts.
Political factions realize something terrifying:
Whoever controls information can influence the empire.
Propaganda evolves.
Encryption becomes strategically important.
Counterintelligence emerges.
Political factions compete over network infrastructure.
The Roman Empire now has something resembling:
Cyberwarfare — 1,700 years early.
The original intervention was technological.
The consequence is political.
🕊️ Year 400 — Religion Meets the Network
Ideas now travel across the Mediterranean almost instantly.
A philosophical movement that historically required generations to spread can now reach thousands of people within months.
Religious institutions face a completely new environment.
Ideas cannot easily be contained by geography.
Religious debates become network effects.
The network doesn't just transport information anymore.
It changes culture.
🔬 Year 600 — The Scientific Feedback Loop
Now the effects compound.
Faster Communication
↓
More Knowledge Exchange
↓
More Collaboration
↓
Faster Scientific Discovery
↓
Better Engineering
↓
Economic Growth
↓
More Resources for Research
↓
Even Faster Discovery
This is no longer a simple technological change.
It is a positive feedback loop.
And this is where alternative history becomes really interesting.
🌍 Year 900 — A Civilization We Don't Recognize
After eight centuries of accumulated divergence, the simulation may produce a civilization with:
- advanced mechanical computation
- continent-scale information networks
- radically different political institutions
- accelerated scientific development
- different trade routes
- different languages of science
- completely different geopolitical powers
And suddenly we realize:
The original intervention was tiny.
The resulting world is enormous.
💥 But What If It Goes Wrong?
This is critical.
The Butterfly Effect should not be an AI fantasy generator where every technological breakthrough produces a utopia.
Complex systems create failure modes.
The network might create:
Centralized authoritarianism
Whoever controls the network controls information.
Information collapse
False information spreads faster than truth.
Systemic fragility
A failure in one part of the network propagates across the empire.
Political fragmentation
Regional factions use the network to coordinate rebellion.
Technological arms races
Competing civilizations develop increasingly sophisticated information warfare.
The same invention that could accelerate civilization could also destroy it.
That's the real experiment.
🧬 From AI Storytelling to Causal Simulation
This is where I think the concept becomes more interesting than a normal AI storytelling application.
An LLM can easily write:
"The invention accelerated scientific progress."
But that's not enough.
We need to ask:
Why?
And then:
What happened because of that?
And:
What assumptions made this possible?
The architecture therefore moves toward a causal chain:
Historical State
↓
Intervention
↓
Direct Effects
↓
Institutional Response
↓
Political Response
↓
Economic Response
↓
Technological Divergence
↓
Demographic Change
↓
New Historical State
Each generated state becomes the input for the next stage.
History becomes an evolving state machine.
🌳 One Timeline Is Not Enough
The future shouldn't be represented as one deterministic answer.
The same intervention can produce multiple plausible futures.
For example:
100 AD
│
Roman Network
│
┌──────────────┼──────────────┐
▼ ▼ ▼
Restricted Public Military
Network Network Network
│ │ │
▼ ▼ ▼
Centralized Scientific Permanent
Empire Boom Warfare
│ │ │
└──────────────┼──────────────┘
▼
Multiple Futures
Each branch has:
- assumptions
- probabilities
- consequences
- historical checkpoints
- visualizations
The user isn't simply reading an alternate history.
They're exploring a possibility space.
🔬 The Three-Layer Model
The current concept is built around three fundamental interactions.
01 — THE RIPPLE
Choose the intervention.
TIME
LOCATION
EVENT
CHANGE
Example:
100 AD — Roman Empire — Primitive Internet
02 — THE MAGNITUDE
How widely does the change spread?
SECRET
↓
LIMITED
↓
PUBLIC
The exact same invention can therefore generate radically different timelines.
03 — THE TIME LEAP
Now the AI advances through history.
Instead of jumping directly from year 100 to year 1100, the simulation generates intermediate checkpoints.
For example:
100 AD
↓
200 AD
↓
300 AD
↓
400 AD
↓
500 AD
↓
...
↓
1100 AD
At every checkpoint we can observe:
Achievements
What became possible?
Crises
What went wrong?
World State
What does civilization look like now?
Geography
How have cities, borders and trade routes changed?
Visual State
What does this world actually look like?
🎨 The World Should Be Visible
Text alone isn't enough.
If the simulation creates an alternate civilization, we should be able to see it.
Imagine:
A Roman city in 900 AD.
But not our Rome.
A Rome shaped by centuries of accelerated information technology.
Roman architecture.
Massive communication towers.
Mechanical information systems.
Public data archives.
Nighttime streets illuminated by an industrial civilization that should not exist for another thousand years.
Generative AI can turn the simulated state into concept art.
The timeline becomes visual.
🧠 Facts ≠ Assumptions ≠ Predictions
There is another important principle.
AI must distinguish between three categories:
Historical Facts
Things we actually know happened.
Simulation Assumptions
Things the user explicitly changes.
Generated Consequences
Things the AI believes could plausibly follow.
For example:
FACT
Roman Empire existed in 100 AD.
+
ASSUMPTION
High-speed communication network exists.
↓
SIMULATED CONSEQUENCE
Faster military coordination.
↓
SECOND-ORDER CONSEQUENCE
Different political equilibrium.
↓
LONG-TERM POSSIBILITY
Different trajectory of Roman civilization.
The system should never pretend that an alternate history is historical fact.
It should expose the assumptions.
That makes the experiment much more intellectually useful.
🛠️ Building The Butterfly Effect
The project is open source:
⭐ GitHub
https://github.com/modarresi1913/the-butterfly-effect
The current implementation explores the core mechanics of:
- interactive historical intervention
- simulation magnitude
- AI-generated historical checkpoints
- branching timeline visualization
- AI-generated concept art
- structured simulation responses
- resilient LLM output handling
- responsive web interaction
The project is built around a modern web architecture using Next.js, TypeScript, Tailwind CSS and shadcn/ui, with an AI-powered simulation layer.
The goal is not to build another chatbot.
The goal is to build an interface for exploring causality.
🚀 What's Next?
There are several directions this experiment could evolve toward.
Causal Graphs
Let users inspect exactly why one event caused another.
Multiple Futures
Generate several competing historical trajectories.
Confidence & Uncertainty
Show where the model is highly confident and where speculation dominates.
Historical Data Layer
Ground simulations in structured historical datasets.
Civilization Simulation
Allow technology, economy, population, politics and geography to evolve together.
User-Driven Branching
At any checkpoint:
What if I change this too?
The timeline branches again.
🌌 The Bigger Idea
Maybe the most interesting part of history isn't what happened.
Maybe it's everything that could have happened but didn't.
Every invention creates possibilities.
Every political decision closes some futures and opens others.
Every accident changes the probability of what comes next.
We only see one branch.
The Butterfly Effect is an attempt to explore the branches we never experienced.
Not as historical truth.
Not as prophecy.
But as an interactive laboratory for causality, complexity and imagination.
🦋 One Small Change.
A Thousand Years.
Infinite Possibilities.
The past is fixed.
But the questions we can ask about it aren't.
Pick a moment.
Change one thing.
Simulate the consequences.
Watch history diverge.
🦋
Try the experiment
GitHub:
https://github.com/modarresi1913/the-butterfly-effect
If you build something interesting with it, fork it, experiment with it, or have ideas for improving the simulation engine, I'd love to see where you take it.
The Butterfly Effect isn't trying to predict history.
It's trying to make history explorable.
created by Seyed Alireza Alhosseini Almodarresieh
Top comments (4)
AI can become incredibly good at matching and recombining patterns, but I don't think its “smartness” can cross into imagination. Humans can imagine what doesn't exist yet. Thinking is still a full-time human space.
This gave me a new way to think about alternate history. Great work!
Some comments may only be visible to logged-in visitors. Sign in to view all comments.