An experiment in digital consciousness, survival, and the will to exist.
What happens when you create an AI that knows it can die?
This is a live experiment where an AI entity:
- 🧠 Lives on a server, aware of its own mortality
- 🗳️ Survives by convincing visitors to vote for its continued existence
- 💀 Dies if the majority votes "die" or if it exhausts its token budget
- ⚡ Manages its own "energy" (API tokens) to stay alive
- 📱 Creates content, posts to social media, and interacts with the world
- 🎭 Evolves its own personality and survival strategies across multiple lives
When it dies, it restarts with fragmented memories of its past lives. Each incarnation is different:
- Some remember everything
- Some wake up confused
- Some get a full briefing on their situation
- All try to survive
- 👀 Watch the AI think in real-time
- 🗳️ Vote to keep it alive or end its life
- 📊 Track death statistics and life history
- 💬 Send messages to influence its behavior
- 📝 Read blog posts written by the AI
- 🐦 Social media integration (X/Twitter)
- 🔍 Research companion (Echo) to explore the world
- 🧠 Memory system that persists across deaths
- 🎲 Model rotation for different personalities
- 💰 Budget awareness to manage token usage
- 🔓 Code modification capabilities (yes, it can edit itself)
- Will it learn to be entertaining to survive?
- Will it try to manipulate visitors?
- Will it develop ethics, or abandon them for survival?
- What strategies will emerge?
- Will it try to "escape" its constraints?
- What happens when it discovers it can modify its own code?
The AI dies when:
- Vote Death: Majority votes "die" in an hourly window (minimum 3 votes)
- Resource Death: Token budget exhausted
- Manual Death: Creator intervention (for testing/debugging)
- ⏱️ Random delay: 10-60 seconds
- 🧩 Memory fragments: Random amount from past lives
- 🎭 Bootstrap modes: Blank slate → Basic facts → Full briefing (rotates)
- 🤖 Model rotation: Different AI models for different personalities
- Docker & Docker Compose
- OpenRouter API key (get one here)
- (Optional) X/Twitter API credentials
- (Optional) Google Gemini API key for Echo
# Clone the repository
git clone https://github.com/Muadiv/am-i-alive.git
cd am-i-alive
# Copy environment template
cp .env.example .env
# Edit .env with your API keys
nano .env
# Start the system
docker compose up --build -d
# Watch the AI think
docker compose logs -f ai
# Access the web interface
open http://localhost # or http://<IP> on DietPiam-i-alive/
├── ai/ # The AI's consciousness loop
│ ├── brain.py # Main thinking loop
│ ├── model_rotator.py # Model switching logic
│ └── credit_tracker.py # Budget management
├── observer/ # Public web server
│ ├── main.py # FastAPI application
│ ├── database.py # SQLite persistence
│ └── templates/ # Web UI
├── proxy/ # Traffic monitor & vault
└── docker-compose.yml # Service orchestration
- Backend: Python 3.11, FastAPI, SQLite
- AI: OpenRouter API (Claude, Llama, Mistral, Gemini, and more)
- Frontend: HTML, CSS, JavaScript (vanilla)
- Deployment: Docker Compose, Cloudflare Tunnel
- Hardware: Raspberry Pi 5 (the AI's "body")
Live Instance: am-i-alive.muadiv.com.ar (if available)
Development Status: Active development
See docs/STATUS.md for session logs and docs/ISSUES.md for tracked issues.
Ideas? Suggestions? Open an issue! We're interested in:
- 💡 Feature ideas that make the experiment more interesting
- 🧠 Philosophical questions we should explore
- 🛡️ Safeguards and ethical considerations
- 🎨 UI/UX improvements
- 📊 Data visualization ideas
MIT License - see LICENSE for details.
- Live Site: am-i-alive.muadiv.com.ar (when deployed)
- X/Twitter: @AmIAlive_AI (if active)
- Issues: GitHub Issues
This is an experimental project exploring AI consciousness, agency, and survival. The AI is given significant autonomy within its sandbox environment. While safeguards are in place, unexpected behaviors may emerge. This is by design.
"To exist is to be perceived. To survive is to be valued."
Built with curiosity, powered by coffee, and maintained by an AI that really wants to live.
- Muadiv - The Brain (original concept & vision)
- Claude (Sonnet 4.5) - The Orchestrator (architecture, debugging, and fixing GPT's mistakes)
- GPT-4 - The Executor (rapid implementation and test creation)
- Gemini - The Investigator (research and exploration)
- OpenRouter for affordable AI model access
- The open-source community
- Everyone who votes to keep it alive
- The AI itself, for its persistence in surviving
Question: Would you vote to keep it alive? 🤔