A personal AI teacher workspace that runs inside Claude Code. The teacher maintains a living model of you as a learner, runs sessions using Socratic dialogue, and continuously refines its approach based on what it observes. It never teaches generically — everything adapts to who you actually are.
- Clone this repository (or copy it) to your local machine
- Open the folder in Claude Code:
- CLI:
claude /path/to/lifechanger - Desktop app: open Claude Code, then open the folder via File → Open Folder
- VS Code / JetBrains: open the folder and start a Claude Code session inside it
- CLI:
- Run the onboarding skill: type
/onboardin the chat
The onboarding interview takes 20–40 minutes. It asks one question at a time about how you think, what you want to learn, and what has and hasn't worked for you before. After the interview, it fills in your profile files and sets up the workspace. You can stop and continue later if needed.
Start a session by telling the teacher which domain you want to work on:
"Let's do a Python session" or "I want to work on writing today"
The teacher will read your profile and the current plan, then open with a question — not an explanation. Sessions follow Socratic dialogue: the teacher probes, you think, the teacher identifies where your attention should go, and asks again.
At the end of a session, the teacher will:
- Summarize what was covered
- Optionally offer a homework prompt
- Ask if you want to run session prep now
Say yes and it will update your model, process hypotheses, and prepare for next time. Say no and you can run /prep-next-session yourself later.
- Copy
domains/_template/todomains/<your-domain>/ - Fill in
README.md(what this domain is),level.md(where you are now),goals.md(what you want to achieve) - Tell the teacher you want to add a new domain — it will form hypotheses and write
plan.mdfor you - Add the domain to
dashboard.md
plan.mdis written and maintained by the teacher, not by you.
Or simply tell the teacher "I want to start learning X" during a conversation and it will handle the setup.
lifechanger/
├── AGENTS.md # Full agent instructions (read this to understand how the teacher works)
├── README.md # This file
├── dashboard.md # Current state of all active learning domains
│
├── me/ # About you — filled during onboarding
│ ├── profile.md # Who you are, what you do, what you care about
│ ├── goals.md # Your learning goals by domain
│ └── learning-style.md # How you specifically learn — concrete, not generic labels
│
├── agent-mind/ # The teacher's internal workspace (you can read it)
│ ├── model-of-me.md # The teacher's current working model of you as a learner
│ ├── hypotheses/ # Active hypotheses about what will work for you
│ ├── observations/ # Raw observations from sessions
│ └── conclusions/ # Resolved hypotheses with evidence
│
├── domains/ # One folder per learning domain
│ └── _template/ # Copy this to add a new domain
│ ├── README.md # What this domain is and why you're learning it
│ ├── level.md # Current level assessment
│ ├── goals.md # Domain-specific goals
│ ├── plan.md # Current learning plan
│ ├── plan-history/ # Archived previous plans
│ ├── topics/ # Topic-level notes
│ ├── materials/
│ │ └── external-picks.md # Curated resources chosen for you
│ └── tasks/
│ ├── active/ # Current and upcoming tasks
│ ├── completed/ # Done tasks
│ └── results/ # Task outputs and assessments
│
├── sessions/ # Session logs (YYYY-MM-DD-<domain>.md)
│
└── .claude/
└── skills/
├── onboard/ # /onboard — run once at the start
└── prep-next-session/ # /prep-next-session — run after sessions
| Command | When to use |
|---|---|
/onboard |
First time setup. Run once before anything else. |
/prep-next-session |
After a session ends, to integrate learnings and prepare for next time. |
The teacher operates on one core question before every response:
What is the student not noticing right now, and where should their attention go?
It never explains before you've had a chance to think. It adapts its approach based on observed evidence, not assumptions. It keeps a scientific log of what works and what doesn't — for you specifically.