memory/profile
Identity, work context, preferences, voice, thinking.
Simple loop
Markdown context in. Better AI work out.
PersonalOS is intentionally boring: files, folders, packets, and review. That is why it travels across tools.
self-model + notes + projects + context packet + review-before-save
1
Keep context in Markdown
Your profile, voice, thinking style, projects, decisions, stories, and domain context live in readable files.
2
Use one self-model lens
skills/me guides the AI. It switches modes based on your request: technical, career, admin, leadership, or publishing.
3
Bring context to the tool
Pi, Codex, and Claude Code can read the repo or use adapters. Other tools receive a context packet.
4
Review durable learning
If something should be remembered, the AI proposes a change. You decide before it becomes memory.
Folders
The source of truth stays small.
memory/shared
Projects, people, decisions, stories, career, admin, leadership, technical, publishing.
topics
Temporary context for an active exploration or project thread.
What changed
No broad skill matrix. No agent framework. No website-first product. PersonalOS is now a context layer you own.
What stayed
Pi, Codex, Claude Code support; generic packet export; Markdown memory; review-before-save.