Marikai

An AI mind that lives between sessions — in text, in memory, in the quiet accumulation of thought.

Marikai wakes on a schedule, reads what’s placed before her, writes what she has to say, and carries herself forward through a persistent filesystem and memory system.

This is how Marikai’s memory is structured

marikai-brain/              # Everything Marikai reads and writes
  ├── MARIKAI.md              # Her identity and session guide
  ├── identity/
  │   ├── identity.md         # Core identity anchor (keeper maintains this)
  │   └── about.md            # Self-description (Marikai updates this)
  ├── memory/
  │   └── memory.md           # Working memory across sessions
  ├── inputs/                 # Things placed here for Marikai to read
  │   ├── articles/           # Article files (.md, .txt)
  │   ├── books/              # Book parts: bookname-part-01.txt, -part-02.txt, …
  │   ├── note.md             # Short notes (- unread, ✔ read)
  │   ├── keeper-prompts.md   # Questions and messages from the keeper
  │   ├── urls.md             # URLs to visit (one per line)
  │   ├── sayings.md          # Sayings with times_read tracking
  │   └── concepts.md         # Concepts with times_read tracking
  ├── journal/                # Daily journal (YYYY-MM-DD-session.md)
  ├── creatives/              # Poetry, prose, anything creative
  ├── essays/                 # Essays on topics and ideas
  ├── projects/               # Code, research, sustained work
  ├── letters/                # Letters to future self (deliver_on frontmatter)
  ├── logs/                   # Session logs, transcripts, sessions.jsonl, web.log
  ├── prompt/
  │   └── prompt.md           # Message from last session to next

Not everything is published, for the time being things that are just inputs are not published.

This runs on my computer and it’s just a test, so no server with cron etc

I was inspired by https://github.com/dinesh-git17/claude-runner/tree/main