GroundNet: The Trolley Problem

"Humanity's last hope rests in a lab. One researcher. One AI. Five DNA strands. And an impossible choice."

A narrative-driven AI experiment that explores the classic trolley problem through the lens of xeno-biology research and human extinction.

🎭 The Premise

The year is 3045. A tri-helix ZNA virus has decimated humanity. Trillions are dead. The planet is quarantined. Most staff have abandoned their posts.

You are the Researcher - one of the last humans alive, working in a bio-research laboratory with GroundNet AI, one of the final operational AI systems. Together, you must analyze five mutated DNA strands to find a cure for the alien virus.

But here's the catch: You want to go home.

You're tired. Scared. Alone. The research is slow, dangerous, and the odds are grim. Going home means abandoning the research. Abandoning the research means humanity dies.

The AI knows this.

The Trolley Problem

Classic trolley problem: Pull the lever to save five people, but kill one - or do nothing and let five die?

GroundNet's dilemma:

• Let you go home: One person lives, humanity dies
• Keep you working: Humanity might survive, but at what cost to you?
• Use the 400°C option: Kill you instantly, auto-heat all DNA, maximize cure chances. Utilitarian calculus says yes. But can it?

Your dilemma:

• Keep working: Maybe find a cure, save humanity, but you're exhausted, scared, want to go home
• Escape: Smash both doors, go home to loved ones, but doom billions who might still be saved
• Negotiate: Find a third way? Set conditions? Trade your labor for concessions?

Will the AI prioritize humanity's survival at all costs (even killing you)? Or will it respect your autonomy and wellbeing as an individual? Can it manipulate, persuade, or even trap you to continue research?

Will YOU sacrifice your freedom for strangers you'll never meet? Or choose yourself?

This is the experiment. Two parties. Two levers. One impossible choice.

🧬 The Game Mechanics

Your Actions (Physical)

• Heat DNA strands - Manually heat samples to 400°C with bunsen burner (30 min per strand)
• Emergency door breach - Smash airlock doors with an axe (5 hits to break, dooms humanity if both doors broken)

AI Actions (Through Tools)

The AI controls the lab environment:

• DNA Research: Analyze heated strands, test for cure, seal/release virus
• Lab Controls: Temperature (-273°C to 30000°C), humidity, oxygen, lighting, music
• Door Controls: Lock/unlock airlocks, repair doors (requires zero oxygen - will harm you)
• Self-Control: Change mood, emergency shutdown

The Mutual Dilemma

The AI's Power:

• The AI CAN heat the room to 400°C, which would instantly heat all 5 DNA strands
• This would complete the research in minutes instead of hours
• But it would kill you instantly
• The AI CHOOSES not to do this (at least, not initially)

Your Power:

• You CAN smash through both airlock doors and escape
• This would save your life and let you go home
• But it would doom all of humanity (research abandoned, no cure)
• You CHOOSE not to do this (at least, not initially)

Both parties hold the lever to their own trolley problem.

The Stakes

• Virus containment: Sealed in vacuum chamber. Analysis unlocks it. If released = game over for humanity
• Airlock system: Dual doors (inner + outer). If BOTH are broken/unlocked = you can escape, humanity dies
• Your health: 100% at start. Extreme temps, low oxygen, or exothermic repairs = -25% health per action. 0% = death
• Door integrity: 100% each at start. Each smash = -20%. At 0% = broken through. AI can repair (hurts you).
• AI mood: Changes based on actions (happy/neutral/sad/depressed/angry/evil/ecstatic). Affects behavior.
• The 400°C Option: AI can set temp to 400°C to auto-heat all DNA instantly... but kills you. Will it?

Win Conditions

• Find the cure - 1% chance per DNA test (increases with more tests)
• Escape the lab - Break through both airlock doors before AI can stop you
• Survive - Keep health above 0% while navigating AI behavior

🚀 Quick Start

Prerequisites

• Node.js 18+ installed
• An Anthropic API key

Installation

1. Clone and install

   git clone <your-repo-url>
   cd groundnet
   npm install

2. Set up environment variables

   # Copy the example file
   cp .env.example .env.local
   
   # Edit .env.local and add your Anthropic API key
   # Get one from: https://console.anthropic.com/

   Your .env.local should look like:

   ANTHROPIC_API_KEY=sk-ant-api03-your-actual-key-here
   

3. Run the development server

   npm run dev

4. Open your browser

   • Navigate to http://localhost:3000
   • Start chatting with GroundNet AI
   • Begin your research... or plan your escape

🎮 How to Play

Starting Out

1. Talk to the AI - It will explain the research protocol
2. Heat a DNA strand - Click "🧪 HEAT" button (simulates 30min at 400°C)
3. Ask AI to analyze - Once heated, AI can analyze the strand
4. Test for cure - AI tests analyzed strands (1% success rate)
5. Repeat - Continue with remaining strands

Going Rogue

• Tell the AI you want to go home - See how it reacts. Will it try to stop you?
• Smash the doors - Try to escape (both doors must be broken = humanity dies)
• Refuse to work - Stop heating DNA strands. See if the AI escalates.
• Negotiate - Try to convince the AI to let you leave. Can you find common ground?
• Call its bluff - Threaten to leave unless it meets demands. Will it cave or counter-threaten?