Here’s How We’ll Learn (Together)

Eight Weekends, Four Milestones, One Real Build

One session each weekend, 3 hours. Four weekends on the technical core of AI products, two running your real PM job on Claude Code, then two on where the field is going and the idea you will build against it. The Tribe runs from Day 1 - live sessions whenever something new lands, monthly offline meetups, and a closed alumni network across 111 companies, for life.

1 · It Stops Being Magic
2 · The Technical Core
3 · You Out-Operate The Room
4 · Your Idea, Picked
Before weekend 1

Pre-read: how LLMs work

Tokens, context windows, what a call costs, and the hard limits behind every AI product decision.

WEEKEND 1

Technical Core

Agents: A Goal, And The Right To Act

Start with a chatbot, end with an agent: the same model, now a loop with a goal and the right to act. Memory, schedules, and where agents go wrong, prompt injection included. You design one for a task you repeat every week.

Due: one agent designed on a page, and three inputs it must survive

WEEKEND 2

Technical Core

MCP: How It Gets Hands

The agent you designed cannot reach anything yet. What the protocol is, why every AI product now speaks it, and which of the tools you live in could be plugged in.

Due: a map of what your agent could plug into, in your own job
1
MILESTONE · AFTER WEEKEND 2

It Stops Being Magic

Tokens, the loop, tools. You can explain what it is doing, out loud.

WEEKEND 3

Technical Core

Evals: Proof Instead Of A Demo

Twenty test cases turn something you hope works into something you would ship. A model as the judge, and reading benchmarks like a PM rather than a marketer.

Due: twenty test cases, five of them run by hand on two models

WEEKEND 4

Technical Core

RAG And Fine-Tuning: Making It Know Your Product

Retrieval over your own documents, what fine-tuning is actually for, and how to tell when a better prompt would have done instead.

Due: the documents it will search, and a call on fine-tuning
2
MILESTONE · AFTER WEEKEND 4

The Technical Core

An agent designed end to end: its tools, its tests and its documents.

Then Claude Code arrives: your Claude Max seat starts before weekend 5, so you can experiment.

WEEKEND 5

Operating System

Your Job, In A Folder

Claude Code becomes the system you work in: your context written down once, your repeat work as saved commands, and the agent you designed, built and scored.

Due: your workspace, and your agent built and passing its tests

WEEKEND 6

Operating System

Run Your Week On It

Research, synthesis and decisions at PM speed, with your agent wired into your real week. Not a demo week. The actual one, with the actual mess in it.

Due: a week of your own work, run on the system
3
MILESTONE · AFTER WEEKEND 6

You Out-Operate The Room

Your job written down once, your agent built, and a real week run on both.

WEEKEND 7

What's Next

What Is Actually Happening Out There

Where the money went, what the platforms absorb, and what is arriving next: models that return a decision instead of text, agents talking to other agents, agents that pay each other. Then your own product, run through the pressure points.

Due: your own product, run through the pressure points

WEEKEND 8

What's Next

How They Actually Won

Nine companies, worked case by case: Chegg priced out of its own market, Intercom charging per resolution instead of per seat, Granola beating fifty rivals on positioning alone, Grammarly buying its way from adjacent tool to workspace. Then the one idea you will build.

Due: one idea, written as a bet
4
MILESTONE · AFTER WEEKEND 8

Your Idea, Picked

One idea, written as a bet. Then twelve months of help building it, on your own clock.

The Job, As It Is Actually Done

Five Decisions, Made For Real

The part slides cannot teach, on a product that is yours. You make all five across the eight weekends and live with what happens.

  1. 01

    What to build

    The one thing it is missing, argued in a room that will tell you if you are wrong.

  2. 02

    What to kill

    Something that works, cut because it is not worth keeping.

  3. 03

    Who it is for

    Twenty named people, and what their objections actually sound like.

  4. 04

    How you know it works

    Twenty test cases, run in front of people. Evidence, not opinion.

  5. 05

    What it costs, and what to charge

    Cost per call, per user, per month, then a price you can defend.

What You Walk Out With

The Six Things That
Are Yours After Eight Weekends

Not slides and a certificate. Three things you make, and three that keep going after the room empties.

A Real Product, Yours To Take To Market

A Real Product, Yours To Take To Market

Bring your own idea, or take one off the shelf that already exists and already works. Either way you leave with the repository, the data, and the right to take it to market.

A Number You Can Defend

A Number You Can Defend

Your own test set, twenty cases, five of them run by hand against two models with cost and latency beside accuracy. You know where it fails, why, and what you would tell your team it cannot be trusted with.

An Operating Setup You Keep

An Operating Setup You Keep

Your job written down once, in a folder Claude Code works from: your context, your repeat work as saved commands, your connectors and your own test set. It is still running the week after the cohort ends.

And after it ends

A Page With Your Name On It

yourname.curious.pm, with what you built on it, the score it earned, and a write-up people outside the room can read.

The Academy, Afterwards

Tutorials, project ladders, the job board, the monthly issue, and coffee in person. Included for as long as you stay engaged.

A Room Of 139 PMs

Across 111 companies and 8 countries, almost never two from the same org. Deep dives every two weeks, meetups in person.

Being Built Right Now

Ideas Already On Their Way. Yours Is Next

  • LIVE

    Podcast Search

    Search hours of audio by what was said, with citations and timestamps.

  • LIVE

    Today Journal

    You call a number, talk, hang up. It remembers across calls.

  • LIVE

    Quiet Humans

    Finds people by what they have built, not what they posted.

  • Yours

    Your idea, picked before the course ends, then built with us.