In classA great product sells itself.
On the jobDistribution decides who wins. Good products die quietly every day because nobody hears about them.
Live cohort · 8 weeks · Free · Cohort 1
AI made building easy. The rest of the job is where products win or die: customers, distribution, metrics, and shipping week after week. Learn it by doing it on your own product, with AI agents as your force multiplier.
Seats are limited to 15 per cohort.
Starts Saturday 12 December. Every Saturday, 10:00 to 12:00 Malaysia time (UTC+8), until 30 January. Add to calendar
Taught by a product leader with 24 years in technology, at
What business school does not teach
In classA great product sells itself.
On the jobDistribution decides who wins. Good products die quietly every day because nobody hears about them.
In classThe idea is the valuable part.
On the jobIdeas are cheap. Execution, every single week, is what compounds.
In classWrite the strategy, then execute the plan.
On the jobYou ship, learn from users and rewrite the plan every few weeks.
In classThe PM writes the specs.
On the jobThe PM owns outcomes: adoption, retention, revenue. Specs are just a tool.
In classFrameworks give you the answer.
On the jobFrameworks sharpen the question. Evidence and judgment answer it.
In classAI will build the product for you.
On the jobAI builds it in days. Getting anyone to use it is still on you.
The new reality
When building took months, most of a PM's energy went into getting things built. Now an agent can build a first version in days.
So the valuable work is everything around it: knowing the customer, choosing what to build, and getting it in front of people. That is where this course spends its time.
Curriculum
Each week: one real decision on your own product, the PM craft behind it, and one new AI agent that takes work off your plate.
Loop 1
Talk to customers, pick one problem, prototype fast.
Weeks 1 to 2
Loop 2
Market, distribution, business model, metrics, then v1.
Weeks 3 to 5
Loop 3
Real users, retention, growth experiments, launch.
Weeks 6 to 8, then Demo Day
Decide: which problem is worth testing first?
Do: write your hypotheses without AI, then interview 5 real people.
AI concept: tokens, context windows, and why models make things up.
Read: The Mom Test (Rob Fitzpatrick); Inspired, part one (Marty Cagan).
Agent: research synthesiser. Turns interview notes into themes, each backed by a real quote.
Decide: which one solution will you test, and what result would kill it?
Do: build a clickable prototype with AI in under an hour, test it with 3 people. First crit.
AI concept: context engineering: what to give the model, and what to leave out.
Read: Continuous Discovery Habits (Teresa Torres), opportunity mapping chapters.
Agent: build kit. A spec and design rules file every AI builder reads first.
Decide: continue, pivot or kill? And which customers first?
Do: hold your gate review against criteria you set in Week 2, then interview 10 people in your segment and find where they already hang out.
AI concept: tools, web search and retrieval; the agent loop.
Read: Crossing the Chasm, chapters 1 to 3 (Geoffrey Moore).
Agent: market watcher. A weekly digest of competitor and market news, with sources.
Decide: where will you win, and how will you make money?
Do: write a two-page strategy memo yourself, then let agents attack it.
AI concept: choosing a model, and using one model to judge another.
Read: 7 Powers, introduction (Hamilton Helmer); Good Strategy Bad Strategy, chapters 1 to 2 (Richard Rumelt).
Agent: red team and judge. Argues against your memo and scores it against your rubric.
Decide: what goes into v1, what stays out, and how will you know it worked?
Do: build v1 with an AI builder that reads your spec, add analytics. Second crit.
AI concept: token costs: cost per task, per user, and how to cut it.
Read: Working Backwards, the PRFAQ chapter (Colin Bryar and Bill Carr).
Agent: metrics brief. Your weekly update: what moved, and what to check.
Decide: what will agents run, and where must you stay in the loop?
Do: plan two weeks of v2 and recruit real users.
AI concept: agent = model + harness; skills, MCP connectors, approvals, test sets.
Read: "Building Effective Agents" (Anthropic); "Beyond the Prompt" (Amit Goel).
Agent: the harness. Your agents connected to your docs, tasks and analytics, tested on 20 real tasks.
Decide: what are users telling you, and what will you change?
Do: get at least 5 real users, watch what they do, ship v2.
AI concept: safety and reliability: prompt injection, private data, regression tests.
Read: Hamel Husain and Shreya Shankar's free AI Evals email course.
Agent: feedback loop. Groups feedback and usage into themes and proposes experiments.
Decide: launch, pivot or kill? Make the call with your data.
Do: write the launch plan and rehearse your Demo Day pitch.
AI concept: explaining AI trade-offs like a product leader: quality, cost, speed, risk.
Agent: interview coach and sceptical investor. Mock interviews and hard pitch questions.
6 minutes each: a live demo, real usage data, your launch decision, and your agents at work. Two guest critics from the industry give feedback, then a career clinic.
Week 0 is setup on your own time. Download the full curriculum (PDF)
Agentic AI, the force multiplier
Interview notes, market scans, metric reports, status updates. By Week 8 a team of agents you built yourself handles them, and each one has a test so you know when to trust it.
You also learn how agents work from first principles, so you can talk tokens, context, cost and evals with engineers and interviewers.
What you leave with
How it runs
Two crits and Demo Day are reviewed by guest product leaders (names announced before the start). Brain first, AI second, evidence third: you think and write before you prompt.
Your coach
Chief Product Officer, Moving Walls · Kuala Lumpur
Amit owns product strategy at Moving Walls, a global out-of-home advertising technology platform, and runs a team building AI agents and MCP servers. Before that he led product at The Trade Desk and Amagi, founded a startup, and spent 11 years as an engineer. He has built his own products too: Penome and Nivritee.
Who it is for
No coding needed. Software, hardware or both.
No fee now and no catch later. If it helps you, help someone else learn.
Free. Seats go in the order people confirm their email. Registration never closes.
Registration opens soon. Check back in a few days, or follow on LinkedIn for the announcement.
Seats are limited to 15 per cohort.
Your details are used only to run this program. Read the privacy notice.
Frequently asked questions
Nothing. The program is free and pro bono, run in a Pay It Forward spirit: if it helps you, help someone else learn.
Every Saturday from 12 December 2026 to 30 January 2027, 10:00 to 12:00 Malaysia time (UTC+8). That is Friday evening in the Americas, very early Saturday in Europe and 07:30 to 09:30 on Saturday in India.
About 6 hours a week for eight weeks: a 2-hour live session plus about 4 hours of your own work (readings, customer interviews, building and your weekly agent).
No. You will build with AI tools that work from plain language. Engineers who want to go further can use coding agents.
No. Claude is recommended and live demonstrations use it, but ChatGPT, Gemini or another assistant works too. The methods are the same.
Yes. You will prototype the app or a simulated version of the device, and work through hardware economics and launch.
The cohort runs as a WhatsApp Community. Official updates go to an announcement group where only the coach and admins can see your number. A separate discussion group is optional; if you join it, members there can see each other's numbers unless you use a WhatsApp username.
Yes. Registration never closes. If Cohort 1 is full when you confirm your email, you keep your place in the queue for the next cohort and are offered any Cohort 1 seat that opens up.
The program gives you something more useful to an employer: a working product, your agent harness with its eval report, a public case study, and written feedback from industry critics.
Eight weeks, live, free. Build a product, get it used, and bring your own AI agents along.
Seats are limited to 15 per cohort.
Last updated