Start Here

How the curriculum works, who it's for, the four-shift operating model, and where to begin.

This curriculum turns the Agentic PM Playbook from a pile of docs into a teachable system. By the end you will run product discovery in a fundamentally different way: outcomes traced to every work item, signals synthesized the same week they arrive, and a team of agents doing the production work while you hold the judgment.

The one thing to internalize first: this is not a course on “how to use plugins.” It is a course on a new operating model. The tooling exists to make that operating model cheap enough to sustain. If you learn the tools without the model, you will just do old-style PM faster — and miss the point.


Before you start: build your desk

If you haven’t set up Obsidian + Claude Threads yet, do that first. The PM Workspace Setup guide walks you through the one-command vault installer, adding your Anthropic API key as a secret, connecting JIRA or Linear, and Vault Bridges — everything you’ll need for this curriculum. (~30 minutes, tool-first, no philosophy.)

Come back here when your workspace is live.


Who this is for

  • PMs adopting the playbook on a real product. Work through the modules in order on the sample dataset, then run the capstone on your own product.
  • Facilitators running a cohort. Each module is a session. The exercises are designed to be done live so you can fix friction on the spot.
  • Self-serve / public adopters. Every module is written to assume zero prior context. If something assumes knowledge you don’t have, that’s a bug — log it (see Troubleshooting below).

You do not need to be technical. Modules 0–4 and 6 require no coding. Module 5 is about delegating technical work to agents, not doing it yourself.


The operating model in one table

From (traditional PM)To (agentic PM)
Output and roadmap featuresOutcomes, with every work item traced to a measurable result
Quarterly research sprintsContinuous weekly discovery; signals processed the same week
PM as document producerPM as judgment holder; agents produce artifacts, PM decides
Binary “ready to build?” gatesProgressive Investment: staged evidence thresholds

Everything in this curriculum serves one of these four shifts.


The learning path

Work the modules in order. Each builds on the artifact the previous one produced — by Module 4 you are operating on a single thread of work (one outcome → one OST → real signals → one tested assumption).

ModuleTitleTimeCoding?Produces
0The Operating ModelHalf dayNoA backward trace from a real feature to its (missing) outcome
1Environment SetupHalf dayA littleA working environment + pm-config.md
2Your First OST1 dayNoA health-checked Opportunity Solution Tree
3Signal Synthesis1 dayNoClustered, evidence-tagged opportunities mapped to the OST
4Experiments & Progressive Investment1 dayNoOne assumption decomposed, leanest test designed, gated
5The Agent Team1 dayDelegation onlyOne solution run from story → design brief
6Cadences & HealthHalf dayNoRecurring rituals scheduled in your own calendar
CapstoneOne Full Cycle on Your Real Product1–2 weeks asyncDependsA complete discovery loop, reviewed against a rubric

The full curriculum — Modules 0–6 and the capstone — is written and linked above. Work them in order; each builds on the artifact the previous one produced.

After the capstone: check the Success Metrics guide to evaluate whether the operating model is actually changing how you work at 30, 60, and 90 days.


How each module is structured

Every module follows the same template (see _module-template.md):

  1. Learning objectives — 3–5 things you’ll be able to do afterward.
  2. Concept reading — a one-page distillation, with a link to the deeper playbook doc. You should never have to read the 25KB master doc on day one.
  3. Hands-on exercise — done on a provided sample dataset (a fake product, fake interviews, fake tickets) so everyone works from the same material.
  4. Success criteria — how you know you did it right.
  5. Common failure modes — the mistakes people actually make, seeded from real use.

The sample dataset

Modules 2–4 operate on a shared fictional product so the exercises are concrete and repeatable. See sample-data/ for the product brief, interview transcripts, support tickets, and a seed OST. (The fictional product is deliberately boring and relatable — a B2B SaaS scheduling tool — so nobody’s real product leaks into the exercises.)

The capstone is the moment you drop the sample data and run the whole loop on your product.


If you get stuck

  • Concept unclear? Each module links its source playbook doc. Read that section, then come back.
  • Tool won’t cooperate? Module 1 is also the install smoke test. Friction there is a real bug — capture it (what you did, what you expected, what happened) so it can be fixed in setup.sh or the docs.
  • Exercise feels wrong? Re-read the Success criteria and Common failure modes before assuming you’re off track. Most “this feels wrong” moments are the intended discomfort of the new operating model.
  • Something specific breaking? The Troubleshooting Guide covers common failure modes across all modules — skill invocation fallbacks, N-sizing for small cohorts, heartbeat automation, and capstone self-scoring.
  • Capstone coming up and want to see a full submission? The worked exemplar is a complete fictional capstone scored against the rubric.

Ready?

Start with Module 0: The Operating Model. Do it before you install anything. The point of Module 0 is to make you slightly uncomfortable about how you work today — that discomfort is what makes the rest of the curriculum stick.