# Personal AI Infrastructure
Personal AI Infrastructure (PAI) is an open-source agentic AI platform by Daniel Miessler built on [[Claude Code]]. Instead of treating AI as a stateless tool, PAI creates a persistent assistant that learns and improves over time; a coach, mentor, and collaborator that knows your goals, beliefs, and projects.
The core thesis: only a fraction of human creative potential is activated, and personal AI infrastructure can change that by making enterprise-grade AI capabilities accessible to everyone.
## Architecture
### TELOS — Deep Goal Understanding
The heart of PAI. Ten documentation files that define who you are and what you're working toward:
Mission, Goals, Projects, Beliefs, Models, Strategies, Narratives, Learned Lessons, Challenges, Ideas.
### Key Systems
- **Skill System**: Modular capabilities that route intelligently based on context
- **Memory System**: Three-tier architecture (hot/warm/cold) enabling continuous learning across sessions
- **Security System**: Permission-aware execution preventing dangerous operations
- **Hook System**: Lifecycle event responses triggering notifications and context loading
- **Voice System**: ElevenLabs-powered text-to-speech with natural prosody
- **Notification System**: Push alerts via ntfy and Discord
- **User/System Separation**: User customizations remain safe during updates
## Guiding Principles
- Scientific method as universal problem-solving
- System architecture matters more than model selection
- Deterministic infrastructure over probabilistic AI
- UNIX philosophy and CLI-first design
- Engineering/SRE principles applied to AI systems
## Tech Stack
[[TypeScript]], Bun, [[Claude Code]]. Installs into `~/.claude/` and runs via the `pai` shell command.
## References
- GitHub: https://github.com/danielmiessler/Personal_AI_Infrastructure
## Related
- [[Claude Code]]
- [[AI Agents]]
- [[Compound Engineering Plugin]]