Personal AI Infrastructure

Magnify what
you're capable of.

PAI is an open-source, AI-based, comprehensive life management system — a dedicated Digital Assistant that moves you from your current state to your ideal state through self-understanding, goal definition, and autonomous, continuous advancement.

PAI startup banner — Kai here, ready to go — with live system stats
pai — the startup banner you see every session
01 — Thesis

AI should magnify everyone — not just the top 1%.

Only a tiny fraction of humanity's creative potential is activated on Earth. Most people don't believe they have valuable contributions to make. They think there are "special" people — and they aren't one of them. Without activation, there is no high-agency. This is the P0 problem PAI exists to solve.

PAI is a Personalized AI Platform designed to magnify your capabilities. It pairs you with a Digital Assistant that knows your goals, preferences, and context — and moves you from your current state toward your ideal state. Not a chatbot. Not a stateless agent. A persistent partner that improves over time.

It's open-source so this quality of AI infrastructure isn't reserved for the rich or technical elite. Everyone gets the best AI.

02Where PAI sits

PAI is the context layer between you and the engines.

At the bottom: AI agent systems like Claude Code, Pi, and OpenCode — the engines that can actually do work. At the top: you — your life, your work, your goals, everything worth managing. PAI sits between them as the context-based life operating system that knows who you are and translates what you want into what the engines should do.

Three-layer PAI architecture: AI agent systems (Claude Code, Pi, OpenCode) at the bottom, PAI context-based life operating system in the middle, and You (life, work, health, goals, relationships, creative, finances, learning) at the top, with context beams flowing upward through all three layers
03The Statusline

Your life, quantified across five dimensions. Every minute.

PAI runs a statusline at the top of every terminal window. Real-time temperature, the five Life OS dimensions (Health, Creativity, Freedom, Relationships, Finances), loaded context files, token burn rate, learning histograms across 6m / 1d / 1mo, and a rotating aphorism. This is what your AI sees when it helps you — and what you see when you open a shell.

HEALTH
sleep · movement
CREATIVE
writing · building
FREEDOM
autonomy · time
RELATIONS
family · friends
FIN
income · runway
PAI statusline showing live health/creative/freedom/relations/fin percentages, context bar, loaded files, usage burn, learning histograms, and a rotating aphorism
pai statusline — live, 60s refresh · terminal-native · always on
04TELOS

Know who you are. Write it down.

TELOS is ten files that capture the whole of you — mission, goals, projects, beliefs, mental models, strategies, narratives, lessons, challenges, and ideas. Your Digital Assistant reads these first. Every suggestion it makes is anchored in what you've already articulated about who you are and what you're working toward.

10 filesmission · goals · projects · beliefs · …
USER/TELOS/
you › Give me an analysis of my current TELOS.
[read] MISSION.md · GOALS.md · PROJECTS.md · BELIEFS.md
[read] MODELS.md · STRATEGIES.md · NARRATIVES.md · LEARNED.md · CHALLENGES.md · IDEAS.md
kai › Updated Telos with P0 activation crisis, P4 corporate opacity, and H3/C3 as solutions.
05Memory

Continuous learning, not stateless chat.

Three-tier memory (hot / warm / cold) with phase-based learning directories. Every interaction emits signals — ratings, sentiment, successes, failures — that feed back into improvement. Corrections become rules. Preferences become defaults. The system stops asking the same questions because it writes down the answers.

3 tiershot · warm · cold · phase-based learning
memory — signal captured
you › Stop asking me — just pick the obvious path.
kai › Sentiment hook works - captured your message just now as rating 3.
kai › Added mandatory immediate feedback to algorithm spec. No more silent hangs.
→ feedback rule saved · MEMORY.md indexed
06Skills

CODE → CLI → PROMPT → SKILL.

A deterministic hierarchy. Code first when you can. CLI tools next. Prompts after. Skills at the top — composable capabilities that route intelligently by context. Research, web design, blog writing, OSINT, security assessment, billing, voice, calendar. All one invocation away.

63+skills shipped in v4 · more every release
skills — real work
you › Redesign PAITUI visual design and explore Electron/Tauri.
[Skill] Webdesign + Explore dispatched
kai › Design gallery ready with 80 concepts. Want me to open it?
kai › Built 23 packs in staging, deployed to PAI repo, awaiting your review.
07Hooks

Eight event types. Your runtime opinions.

Hooks respond to lifecycle events — session start, tool use, prompt submit, stop, notification, pre-compact, post-compact, session end. Validate. Inject context. Notify. Log. Block. The system behaves the way you want, not the way a vendor decided.

8 eventsSessionStart · PreToolUse · PostToolUse · …
PreToolUse — ContainmentGuard
PreToolUse: Write src/components/Footer.tsx
[ContainmentGuard] scanning content
🚨 BLOCKED: private reference outside Z1–Z4 zones
kai › Rerouted. Footer now references only public repo.
08Agents

Persistent assistant. Parallel specialists.

PAI treats AI as a persistent assistant — friend, coach, mentor — not a stateless agent. When depth is needed, it dispatches specialists in parallel: engineers, architects, researchers, QA. Each with its own personality and voice. Each isolated, scoped, and background by default.

14+agent personalities · composable teams
Task — parallel research
you › Do extensive research on Claude Code team members on X.
[Task] spawning 4 researchers: Claude · Gemini · Perplexity · Grok
kai › Found thirteen Claude Code team members on X. Boris, Cat, and Ado are most active.
kai › Grok found Erik Schluntz and Jeremy Fox. Total fifteen handles confirmed.
09Voice

Your AI has a voice.

ElevenLabs TTS on every response. Named character voices per agent. Prosody enhancement so speech sounds natural — not the default robot. Hear task completions, session summaries, important updates spoken aloud while your eyes stay on the work.

ElevenLabsprosody-enhanced · per-agent voices
voice lane — Kai speaking
[voice] Removed text-cleaner layer. Voice server handles all sanitization now.
[voice] Added version and component counts to statusline second line.
[voice] Cognitive Progress Loop diagram complete. Ten stages with central insight callout.
[voice] The Last Algorithm is now live on your blog.
10Primitives

What PAI is made of.

A small set of composable primitives, each you can read, fork, or replace. No black boxes. No vendor lock.

Foundation
  • Claude Code
  • TypeScript
  • Bun
  • Bash
Identity
  • TELOS (10 files)
  • USER/SYSTEM separation
  • upgrade-safe customization
Execution
  • Skills (63+)
  • Hooks (21)
  • Agents (14+)
  • Workflows (180+)
Memory
  • 3-tier hot/warm/cold
  • phase-based learning
  • continuous signal capture
Surfaces
  • CLI
  • Terminal statusline
  • Voice (ElevenLabs)
  • Notifications (ntfy · Discord)
Security
  • policy hooks
  • no --dangerously-skip-permissions
  • containment zones
Installation
  • AI-based GUI installer
  • platform detection
  • identity wizard
License
  • MIT
  • open-source
  • self-hosted
11Install

Clone it. Run the installer. You're running PAI.

The GUI installer detects your system, installs prerequisites (Bun, Git, Claude Code), asks for your identity and preferences, and configures the repo into ~/.claude/. USER/ files are preserved on every upgrade.

install — v4.0.3
$ git clone https://github.com/danielmiessler/Personal_AI_Infrastructure.git
$ cd Personal_AI_Infrastructure/Releases/v4.0.3
$ cp -r .claude ~/ && cd ~/.claude && bash install.sh
$ source ~/.zshrc && pai

Upgrading from an earlier version? The installer auto-detects existing installs, merges settings non-destructively, and never overwrites your USER/ files or custom hooks.

12Community

People are building this with us.

PAI is open-source on GitHub. Star it, fork it, open an issue, send a PR. The list below is live — pulled from the repo right now.

Stars
Forks
Open issues