An adaptive AI coaching system that uses spaced repetition, 10 teaching approaches, and expert personas to make you a world-class Product Manager.
Architected by Alex Wirtzer. Got feedback? Send it to Genie.Wirtzer@gmail.com and we will look into it!
PM Coach is a complete coaching system that turns your AI coding assistant into a persistent, adaptive PM tutor. Unlike asking ChatGPT to "help me prepare for PM interviews," PM Coach remembers everything — your strengths, your gaps, which teaching techniques work for you, and exactly when you need to review each topic.
The system tracks your progress across 13 learning tracks organized into 8 pillars of PM excellence. It uses a spaced repetition algorithm (SM-2 variant) to schedule reviews at optimal intervals, rotates through 10 pedagogical approaches with automatic technique retirement when something isn't working, and tests your conversational readiness through 8 expert persona simulations — from a "VP of Research at DeepMind" probing your AI knowledge to a "Principal Engineer pushing back" on your product decisions.
Everything is stored in plain text files: JSON for state, JSONL for history, Markdown for lesson plans and reflections. You can read every piece of your learning state, version-control your progress, and customize anything. There's no server, no database, no subscription — just you, your AI, and a well-structured file system.
git clone https://github.com/Wirtzer/Product-Management-Coach.git
cd Product-Management-Coach./setup.shThe interactive wizard will configure your profile, check dependencies, and initialize the system.
With Claude Code (recommended):
claude
# Then say: "let's go"With Cursor:
Open the pm-coach folder in Cursor. The .cursorrules file loads automatically. Start a new chat and say "let's go."
With Windsurf:
Copy .cursorrules to .windsurfrules, then open the folder in Windsurf.
The coach will:
- Read your profile and current track state
- Identify what's due for review (on first run: everything)
- Start with your weakest area or a framework introduction
- Score your responses and update your tracks
- Plan the next session
Asking ChatGPT "Help me prepare for PM interviews" gets you a generic answer. Ask again tomorrow and it has no memory of yesterday. You're starting from zero every time.
PM skills are built through deliberate practice with feedback loops — the same way musicians practice scales or athletes drill fundamentals. You need:
- Persistence: Track what you've practiced and how you scored
- Spacing: Review topics at optimal intervals for long-term retention
- Adaptation: Retire techniques that aren't working, intensify what is
- Pressure testing: Simulate real conversations, not just Q&A
PM Coach provides all four.
| Feature | ChatGPT / Generic AI | PM Coach |
|---|---|---|
| Remembers past sessions | ❌ Starts fresh each time | ✅ 13 tracks with persistent scores and full history |
| Adapts teaching approach | ❌ Same approach every time | ✅ 10 pedagogy approaches with automatic technique rotation |
| Spaced repetition | ❌ No scheduling | ✅ SM-2 variant with per-track intervals (1-30 days) |
| Expert persona tests | ❌ Generic role-play | ✅ 8 calibrated expert personas with scoring rubrics |
| Automated reflection | ❌ No analysis loop | ✅ Nightly reflection analyzes patterns and updates plans |
| PM-specific frameworks | Generic knowledge | ✅ Deep framework library: CIRCLES, RICE, JTBD, Working Backwards, and more |
| Progress tracking | None | ✅ Dashboard with 8 pillars, per-dimension scoring, trend analysis |
| Knowledge pipeline | Manual | ✅ Feed articles and papers → auto-scored and routed to tracks |
| Personalized to your background | ❌ No context about you | ✅ Portfolio synthesis: resume, projects, wins inform every session |
| Add your own interview questions | N/A | ✅ Structured question bank with company tags and scoring |
┌──────────────────────────────────────────────────┐
│ YOUR COACHING SESSION │
│ │
│ You ←→ LLM ←→ CLAUDE.md (system prompt) │
│ │ │
│ ├─ reads track state + lesson plans │
│ ├─ reads pedagogy approaches │
│ ├─ reads memory + frameworks │
│ │ │
│ └─ writes updated scores + history │
└──────────────────────┬───────────────────────────┘
│
┌────────────┼────────────┐
▼ ▼ ▼
┌─────────────┐ ┌────────┐ ┌──────────┐
│ Reflection │ │Session │ │Knowledge │
│ Engine (LLM) │ │Prep │ │Pipeline │
└──────┬──────┘ └───┬────┘ └────┬─────┘
│ │ │
▼ ▼ ▼
┌──────────────────────────────────────┐
│ FLAT FILE STATE │
│ tracks/ → JSON scores + JSONL logs │
│ pedagogy/ → teaching methodology │
│ reflections/ → analysis archive │
│ dashboard.md → progress overview │
│ memory.md → session context │
└──────────────────────────────────────┘
All state is in human-readable flat files. No database. Git-friendly. See docs/ARCHITECTURE.md for the full technical deep-dive.
PM Coach organizes learning into 8 pillars covering the full spectrum of PM excellence:
| # | Pillar | Tracks | What You'll Learn |
|---|---|---|---|
| 1 | 🔬 Technical Depth | AI Foundations, World Models, Agent Architectures, Robot Learning | AI/ML concepts, model architectures, research landscape |
| 2 | 🛠️ PM Craft | Product Sense, PM Frameworks | Product intuition, CIRCLES, RICE, JTBD, metrics design |
| 3 | 🎯 Interview Excellence | LP Stories | STAR stories, behavioral answers, delivery under pressure |
| 4 | 🔭 Strategic Vision | AI Product Strategy | Industry landscape, product strategy, market analysis |
| 5 | ✍️ Narrative & Influence | Narrative & Influence | Product narratives, strategy memos, stakeholder communication |
| 6 | 🕸️ Network Intelligence | Network Intelligence, Industry Landscape | Talent flows, company dynamics, ecosystem mapping |
| 7 | 🔨 Builder's Credibility | Builder's Credibility | Technical specs, architecture review, prototyping |
| 8 | 🧠 Decision Science | Decision Science | Cognitive biases, decision frameworks, probabilistic thinking |
Each pillar contains 1-4 tracks. Each track has its own mastery scores, conversational readiness scores, spaced repetition schedule, and lesson plan.
PM Coach uses a 5-tier quality system adapted from the SM-2 algorithm:
| Quality | Score Range | Interval Change |
|---|---|---|
| 5 — Perfect | ≥ 90 | × 2.5 (max 30d) |
| 4 — Good | 80–89 | × 2.0 (max 30d) |
| 3 — Adequate | 70–79 | × 1.5 (max 30d) |
| 2 — Poor | 55–69 | Reset to 1 day |
| 1 — Fail | < 55 | Reset to 1 day + flag |
Tracks you master get reviewed monthly. Tracks you struggle with appear daily until they stabilize.
Each approach has a 5-phase progression with specific techniques per phase:
- Storytelling Practice — Structure and deliver STAR stories fluently
- Framework Application — Internalize PM frameworks for automatic deployment
- Case Analysis — Pattern recognition and structured analytical thinking
- Concept Learning — Deep understanding through concrete-to-abstract progression
- AI Concept Learning — Technical AI depth for expert-level conversation
- Industry Case Analysis — Current, deep knowledge of the AI landscape
- Interview Drill — Integration of all skills under realistic pressure
- Craft-Based Practice — Persuasive communication through iterative writing
- Pattern Recognition & Briefing — Industry awareness and intelligence synthesis
- Project-Based Learning — Technical credibility through hands-on work
The reflection engine monitors which techniques work for you and automatically retires low-performing ones after 3+ sessions.
8 domain expert personas test your conversational readiness — your ability to discuss topics naturally, not just answer quiz questions:
| Persona | Tests | Key Emphasis |
|---|---|---|
| 🔬 VP of Research (DeepMind) | Technical depth | Accuracy, depth, current awareness |
| 🛠️ Principal Engineer | PM craft | Depth, opinion formation, accuracy |
| 🎯 Bar Raiser | Interview skills | Accuracy, fluency, depth |
| 🔭 CPO (frontier AI company) | Strategic vision | Opinion formation, current awareness |
| ✍️ VP of Product | Narrative & influence | Opinion formation, fluency |
| 🕸️ Conference Attendee | Network intelligence | Current awareness, opinion formation |
| 🔨 Hiring Manager | Builder's credibility | Accuracy, depth, builder fluency |
| 🧠 Board Member | Decision science | Depth, opinion formation, accuracy |
Each test is scored on 5 dimensions (100 points total) and updates your conversational readiness scores.
You advance to the next phase when all three criteria are met:
- Quantitative: Mastery score for the current phase's target dimensions exceeds 75/100
- Qualitative: You demonstrate the phase's mastery signals in 2+ consecutive sessions
- Spaced Repetition: At least one successful review after a 3+ day gap
- Adaptive Coaching — Teaching approach evolves based on what works for you
- Mock Interviews — Company-specific with structured STAR feedback
- Framework Teaching — Deep dives with worked examples and practice scenarios
- Expert Tests — Simulated conversations with domain experts who score you
- Daily Reflection — Automated analysis of progress, technique effectiveness, and lesson planning
- Knowledge Pipeline — Feed articles, papers, and notes → auto-scored and integrated into tracks
- Progress Dashboard — Visual tracking across all 8 pillars with ASCII progress bars
- Spaced Repetition — Never forget what you've learned; review at optimal intervals
- Personal Portfolio — Drop your resume, project write-ups, and win lists for deeply personalized coaching
- Question Bank — Add your own interview questions following the included template
- PM Framework Library — 10 reference guides for core PM frameworks (CIRCLES, RICE, JTBD, and more)
pm-coach/
├── CLAUDE.md # System prompt (the coach's brain)
├── .cursorrules # Cursor/Windsurf rules (compact CLAUDE.md)
├── setup.sh # Interactive setup wizard
├── config.json # Your profile and preferences
├── memory.md # Persistent session memory
│
├── learning/
│ ├── dashboard.md # Progress dashboard (auto-generated)
│ ├── session-prep.md # Today's session plan (auto-generated)
│ ├── knowledge-queue.jsonl # Knowledge pipeline queue
│ ├── tracks/
│ │ ├── ai-foundations/
│ │ │ ├── track.json # Scores, spaced repetition state
│ │ │ ├── history.jsonl # Session-by-session log
│ │ │ └── lesson-plan.md # Current teaching plan
│ │ ├── pm-frameworks/
│ │ ├── lp-stories/
│ │ └── ... (13 tracks total)
│ ├── pedagogy/
│ │ ├── approaches.md # 10 teaching approaches
│ │ └── expert-personas.md # 8 expert personas
│ └── reflections/
│ ├── latest.md # Most recent reflection
│ └── archive/ # Historical reflections
│
├── data/
│ ├── frameworks/ # PM framework reference docs
│ │ ├── circles.md
│ │ ├── rice.md
│ │ ├── working-backwards.md
│ │ ├── jobs-to-be-done.md
│ │ ├── north-star-metric.md
│ │ ├── okrs.md
│ │ ├── star-soar.md
│ │ ├── first-principles.md
│ │ ├── opportunity-solution-tree.md
│ │ └── kano-model.md
│ ├── question-bank/ # Your interview questions (add your own)
│ │ ├── README.md
│ │ └── template.json
│ └── portfolio/ # Your personal docs (resume, projects, wins)
│ ├── README.md
│ └── profile-synthesis.md # Coach-maintained synthesized profile
│
├── knowledge/ # Drop articles and papers here
│
├── scripts/
│ ├── update-dashboard.sh # Regenerate dashboard from track data
│ ├── prepare-session.sh # Generate session prep for today
│ ├── run-reflection.sh # Run the nightly reflection analysis
│ ├── feed-knowledge-queue.sh # Process knowledge sources into queue
│ └── add-topic.sh # "Teach me about X"
│
├── docs/
│ ├── ARCHITECTURE.md # System architecture deep-dive
│ ├── CUSTOMIZATION.md # How to personalize everything
│ ├── PLATFORMS.md # Setup for Claude Code, Cursor, Windsurf
│ ├── CONTRIBUTING.md # How to contribute
│ └── FAQ.md # Common questions
│
└── examples/
├── session-transcript.md # Annotated coaching session
├── expert-test-example.md # Annotated expert test
└── custom-track-example/ # How to create a custom track
PM Coach includes a framework library with original syntheses of 10 core PM frameworks, each with worked examples and interview application guidance.
The knowledge/ directory is designed for you to add your own learning materials:
- Articles and blog posts — Drop
.mdfiles and runfeed-knowledge-queue.sh - Research paper summaries — Markdown summaries of relevant papers
- Podcast notes — We recommend distilling insights from Lenny's Podcast by Lenny Rachitsky — organize notes by PM topic (product sense, metrics, growth, strategy, etc.) and drop them in
knowledge/ - Your own notes — Anything relevant to your PM growth
The knowledge pipeline automatically scores new materials for relevance and routes them to the appropriate learning tracks.
| Platform | Status | How It Works |
|---|---|---|
| Claude Code | ✅ Full support | CLAUDE.md auto-loads as system prompt |
| Cursor | ✅ Full support | .cursorrules auto-loads |
| Windsurf | ✅ Full support | Copy .cursorrules → .windsurfrules |
| Aider | Needs manual system prompt loading | |
| ChatGPT | No file persistence or automation | |
| Other LLMs | Needs tool use + file access capability |
See docs/PLATFORMS.md for detailed setup instructions per platform.
All scripts require only bash and jq. No other dependencies.
The reflection engine analyzes your progress, updates lesson plans, rotates techniques, and processes the knowledge queue:
./scripts/run-reflection.shOptional: Run nightly via cron:
# Edit crontab
crontab -e
# Add (runs at 11 PM daily):
0 23 * * * cd /path/to/pm-coach && ./scripts/run-reflection.sh >> /tmp/pm-coach-reflection.log 2>&1Generates a session plan based on what's due and what needs attention:
./scripts/prepare-session.sh# Regenerate the progress dashboard
./scripts/update-dashboard.sh
# Add a topic to learn
./scripts/add-topic.sh "transformer attention mechanisms"
./scripts/add-topic.sh "RICE prioritization" --tracks pm-frameworks
# Process new knowledge sources
./scripts/feed-knowledge-queue.shPM Coach is designed to be extended. Common customizations:
- Add a new learning track — Create a directory in
learning/tracks/withtrack.json,history.jsonl, andlesson-plan.md - Add your portfolio — Drop resume, project write-ups, win lists, and performance reviews into
data/portfolio/for personalized coaching - Add interview questions — Follow the template in
data/question-bank/ - Add knowledge — Drop
.mdfiles inknowledge/and run the pipeline - Modify expert personas — Edit
learning/pedagogy/expert-personas.md - Change the LLM model — Set
REFLECTION_MODELenv var or editconfig.json - Adapt for non-PM domains — The architecture is domain-agnostic; swap the content
See docs/CUSTOMIZATION.md for detailed guides.
We welcome contributions! Whether it's new frameworks, better pedagogy approaches, additional expert personas, or script improvements.
See docs/CONTRIBUTING.md for guidelines.
MIT + Commons Clause — see LICENSE.
Free to use, modify, and fork. The Commons Clause restricts selling the software or a competing product/service built on it without written permission from Alex Wirtzer.
- Lenny's Podcast by Lenny Rachitsky — PM insights distilled from this essential podcast for product people informed many of the framework applications and coaching patterns in this system
- Product Manager Skills by Dean Peters — Complementary PM skills library
- Spaced repetition algorithm adapted from SM-2 by Piotr Wozniak
- Coaching methodology informed by The Prompt Report (2025) — a comprehensive survey of prompting techniques
- PM frameworks synthesized from publicly available industry-standard sources (CIRCLES, RICE, JTBD, OKRs, Kano, and more)