Skip to content

SETUP-GUIDE.md

Latest commit

 

History

History
153 lines (113 loc) · 6.09 KB

SETUP-GUIDE.md

File metadata and controls

153 lines (113 loc) · 6.09 KB
Raw

Setup Guide

Step-by-step instructions to create your own AI-powered curriculum using the Adaptive Learning Engine v4.

Step 1: Fork or Clone This Repo

git clone https://github.com/hanselhansel/ai-learning-engine.git my-learning-project
cd my-learning-project

Or click "Use this template" on GitHub.

Step 2: Design Your Curriculum

Open CURRICULUM-TEMPLATE.md and fill it in:

  1. Define your goal. What does "done" look like? Be specific: "Get hired as X", "Ship project Y", "Pass certification Z".
  2. Set the timeline. How many weeks? How many hours per day? This determines your total session count.
  3. Define phases. Recommended structure:
    • Phase 1: Foundations (build mental model)
    • Phase 2: Technical Literacy (go deeper)
    • Phase 3: Applied Knowledge (build something)
    • Phase 4: Portfolio + Networking (demonstrate and connect)
  4. Write day-by-day content. For each day, include:
    • Core question (the one thing this day answers)
    • Topics table (max 7 concepts per day — the engine splits larger days across sessions)
    • Resources
    • Exercise (bridges to existing knowledge)
  5. Define phase gates. What does the learner need to demonstrate before advancing?
  6. List terminology. Group terms into Tier 1 (must-know) and Tier 2 (know by end of phase).

Save as CURRICULUM.md.

Step 2.5: Set Your Learner Profile (NEW in v4)

If you're using /generate-curriculum, this is handled automatically during the 8-question setup. If you're configuring manually:

  1. Learner Profile: In CURRICULUM-CONFIG.md, fill in the Learner Profile section:

    • age_group: Your age range (this affects analogies and vocabulary)
    • experience_level: How much you know about the topic
    • persona: Auto-assigned, or manually set to Elementary / Teen / Adult Beginner / Professional / Expert

  2. Learning Objective: Fill in the Learning Objective section:

    • primary_goal: Why you're learning this (career, academic, investing, hobby, teaching, building)
    • career_target: Your specific end goal
    • perspective_lenses: The angles you want threaded through every lesson
    • objective_statement: One sentence summarizing your why

  3. Placement Test (optional): Run /generate-curriculum even for an existing curriculum — it will detect the existing files and offer to run just the placement test to auto-assign your persona.

The engine adapts its teaching style (vocabulary, analogies, question difficulty, feedback tone) based on your persona, and threads your learning goal through every question and exercise.

Step 3: Configure the Engine

Open CURRICULUM-CONFIG-TEMPLATE.md and customize:

  1. Curriculum Identity: Set file paths to match your folder structure
  2. Learner Bridges: List your existing expertise that connects to new material
  3. Exercise Types: Define the portfolio artifact formats for your field
  4. Mastery Thresholds: Set per-phase pass/fail scores
  5. Tone & Style: How you want the AI to communicate

Save as CURRICULUM-CONFIG.md.

Step 4: Set Up Session State

Open SESSION-STATE.md and customize:

  • Replace [YOUR TOPIC] with your topic
  • Set [TARGET_SESSIONS] (estimate: ~1.3x your curriculum day count, since multi-concept days split)
  • Update Phase names in Assessment Scores table
  • Leave the SM-2 tracker empty — the engine fills it during sessions

Step 5: Set Up Communities (Optional)

Open COMMUNITIES-TEMPLATE.md and:

  • Research 8-10 communities in your field (Discord, Reddit, X/Twitter, newsletters)
  • Fill in the tables
  • Save as COMMUNITIES.md

Step 6: Install the Skills

Option A: Claude Code / Cowork (Recommended)

Copy the skill folders into your Claude skills directory:

cp -r skills/learn ~/.claude/skills/learn
cp -r skills/learn-end ~/.claude/skills/learn-end

Option B: CLAUDE.md Integration

If you prefer natural language triggers:

  1. Open CLAUDE-SNIPPET.md
  2. Copy the content
  3. Paste into your project's CLAUDE.md file
  4. Replace all [BRACKETED] values with your paths

Step 7: Populate Research Reference (Optional)

Open RESEARCH-REFERENCE.md and:

  • Run web searches for your topic's latest data
  • Fill in companies, metrics, papers, events
  • Include timestamps for all data points

The engine uses this file as source of truth during sessions.

Step 8: Commit and Start Learning

git add .
git commit -m "Initialize curriculum: [your topic]"
git push

Then open Claude and type: "let's learn"

Folder Structure After Setup

your-curriculum/
├── CURRICULUM.md              <- Your filled-in curriculum
├── CURRICULUM-CONFIG.md       <- Engine configuration + learner profile + objectives
├── SESSION-STATE.md           <- Session state (auto-managed)
├── COMMUNITIES.md             <- Your community tracker
├── RESEARCH-REFERENCE.md      <- Verified data for your topic
├── progress.md                <- Session log (auto-updated)
├── skills/
│   ├── learn/SKILL.md         <- Adaptive learning engine
│   └── learn-end/SKILL.md    <- Session exit handler
├── portfolio/                 <- Portfolio artifacts + analytics dashboard + showcase
├── sessions/
│   ├── summaries/             <- Per-session receipt cards
│   ├── digests/               <- Weekly learning digests
│   └── archive/               <- Archived state history
├── assessments/               <- Phase gate results
├── swot/                      <- Self-assessments
├── build-project/             <- Portfolio project files
└── outreach/                  <- Networking tracker

Tips

  • Don't over-plan. Write Phase 1 in detail, outline Phase 2-4 at a higher level, refine as you learn.
  • Use the bridges. Every concept should connect to something you already know. Write these into CURRICULUM-CONFIG.md.
  • Trust the state machine. Don't manually edit SESSION-STATE.md unless something breaks.
  • Commit often. Your learning history is a portfolio piece.
  • Do multiple sessions. Sessions are decoupled from calendar days. Go as fast as you want.