Personal AI Infrastructure (PAI), now LifeOS
Caution
Renamed. The project described here as PAI (Personal AI Infrastructure) was renamed LifeOS with release v6.0.0 on 2026-07-02 (“the project was PAI; this is the same system, renamed”). The GitHub repository
danielmiessler/Personal_AI_Infrastructurenow redirects todanielmiessler/LifeOS. This note keeps the PAI filename and aliases because video titles from 2025 to mid-2026 use “PAI” and “Personal AI Infrastructure”.
LifeOS / PAI is an open-source (MIT) “AI harness” by Daniel Miessler that layers skills, hooks, agents, memory and a personal-context model (TELOS) on top of an AI coding harness, primarily Claude Code, so the assistant knows the user’s goals and improves over time. Repo: https://github.com/danielmiessler/LifeOS. Written mostly in TypeScript and run with Bun; install is done by the user’s AI (prompt “Read https://ourlifeos.ai/install and install LifeOS for me”) or a curl script for Claude Code on macOS/Linux.
Version timeline (GitHub releases, fetched 2026-10-05)
| Date | Release | What changed |
|---|---|---|
| 2025-09-21 | v0.1 | first release (early v0.x series) |
| 2025-10-17 | v0.4.0 | repository restructure (breaking) |
| 2025-12-30 | v2.0.0 | ”PAI Packs”: modular, AI-installable capability bundles; designed to work with Claude Code, OpenCode, Gemini Code, GPT-Codex or custom systems |
| 2026-02-15 | v3.0.0 | ”The Algorithm Matures”: constraint extraction, installer, loop mode with parallel workers |
| 2026-02-28 | v4.0.0 | 38 flat skill directories compressed into 12 categories |
| 2026-05-13 | v5.0.0 | ”Life Operating System”: daemon, Life Dashboard, personalized Digital Assistant |
| 2026-07-02 | v6.0.0 | first release as LifeOS; ships as one self-contained skill (49 skills at that release) |
| 2026-07-12 | v7.0.0 | ”The Bitter Pill Release” |
| 2026-08-14 | v7.40.4 | latest release seen (as of 2026-10-05 the repo had ~19.3k stars) |
The older v0.9.0 “Platform Agnostic Release” figures in earlier versions of this note (65+ skills, 227+ commits) are obsolete.
Core concepts (as described in the project README)
- Skills: self-contained domain capabilities (research, security, writing, art and more), bundled in one skill since v6.
- Hooks: event-driven automation around the harness lifecycle.
- Agents and the Digital Assistant (DA): an identity layer (principal plus DA identity) loaded at session start;
/interviewsets up name, voice and TELOS. - TELOS: the user’s goals/mission context. The Algorithm: a phased “current state to ideal state” execution loop.
- Memory / history and Pulse: captured work history, plus a Life Dashboard and menu-bar app (“Pulse”).
- Voice and a security layer are listed among the components.
- Philosophy, as stated by the author: scaffolding matters more than the model; start small; reliability before features; text-first. Miessler’s longer essay: https://danielmiessler.com/blog/personal-ai-infrastructure.
Requirements and caveats
- Needs a capable AI coding harness (built and run on Claude Code, so a Claude subscription or API access) and Bun.
- The release timeline above shows a rename and a repository restructure (breaking changes); (opinion) expect maintenance work on customizations.
- Self-description is the main source; independent reviews were not collected (see Open items).
Related notes
Claude Code, personal-ai-systems, Second brain with AI, Model Context Protocol.
Sources
- https://github.com/danielmiessler/LifeOS and its README and releases API (fetched 2026-10-05)
- https://raw.githubusercontent.com/danielmiessler/Personal_AI_Infrastructure/main/README.md (redirects to LifeOS; fetched 2026-10-05)
Open items
- Daniel Miessler has no person note in RESOURCES/PEOPLE; his Live Stream is in YT-VIDEO/RSS.
- Independent evaluation and adoption numbers (beyond GitHub stars) not collected.
- Whether the “Kai” name for his personal DA is still used in v7 not checked.