Live training for Product Managers, Senior PMs and Technical PMs
Build and control real AI agents in plain text
Run live agents in your first 15 minutes. Then design your own with Raw Agent Control (RAC). No code.
- 15 minutes to your first live agent
- 100% browser lab
- Nothing to install
- Live agents
- Plain-text control
- Tools and MCP
- Memory
- Multi-agent handoffs
- LLM as a judge
- Human in the loop
- Guardrails
- Browser lab
The problem
Love AI agents, but tired of results you can't fix?
Most PMs watch agent demos. Very few can open the box and change what the agent does.
- Black boxesNo-code tools you can't see inside, and demos engineers won't trust.
- Slow changesTwo sprints to change one prompt or one retry rule.
- Silent failuresAgents that loop, forget or invent data when nobody is watching.
Our belief
Programmers own the syntax. Product managers own the details: the users, the edge cases, the boundaries. In AI agents, the details are the product.
How it works
Three steps from watching to building
01Run live agents
02Master RAC
03Become Agentsable
01
Run live agents
In AI agents, the details matter more than the code. PMs own the details.
In your first 15 minutes you run real agents in a browser lab. You watch them think, fail and respond to your plain-text changes, live. Slides can't teach this.
02
Master Raw Agent Control
RAC is a plain-text way to design agent systems. Like Markdown, but for agents.
Define agents, tools, guardrails and pass-or-repeat gates in structured text files. Test them live, then hand the same files to your engineers. No black boxes, no framework lock-in.
03
Become Agentsable
Turn the ideas in your head into working AI systems.
You leave having built, debugged and refined a real multi-agent system yourself. The gap between a breakthrough idea and a working system gets very small.
Every week
One week, four steps
Built for busy PMs. Miss a video and you still keep up: the live sessions cover it.
Watch the lesson
Short recorded videos, on your own time.
Join the live Q&A
A quick recap and your questions, on Zoom.
Watch a live build
A complete agent system, built in plain text.
Build it yourself
Run and change the same system in your browser lab.
Step 4 in action: you take the live build and steer it yourself.
The four weeks
From fragile prompts to production control
Each week adds one layer of control. Each ends with a hands-on lab.
- 01
Week 1 · Agent brains
From prompt to agent
Turn a fragile prompt into a structured agent.
Lab: Take over the agent from the live build that is misclassifying edge cases. Fix its reasoning behavior by refining its plain text RAC rules.
- 02
Week 2 · Tools & memory
Tools, MCP & memory
Give agents tools, memory and clean handoffs.
Lab: Take the multi agent workflow from the live build and wire its tools, memory rules, and agent to agent handoffs in plain text so the team completes a multi step task end to end.
- 03
Week 3 · Failure patterns
When agents fail
Break agents on purpose, then fix them.
Lab: Inspect a broken multi agent workflow trace from the live build. Diagnose the exact failure pattern and stop the loop by adjusting the plain text definitions.
- 04
Week 4 · Production control
Production control
Add guardrails, judges and human approval.
Lab: Add plain text guardrails, an LLM as a Judge pass or repeat gate, and a Human in the Loop escalation rule to the live build example. Practise the controls needed for production readiness.
Why agents fail
It works on Friday in front of the VP. On Monday it meets real data.
Loops, forgotten rules, invented data, one bad step poisoning the rest. We trigger these failures on purpose, so you learn to fix them.
The four failure patterns
Pick one. Watch it break. Watch RAC fix it.
The agent repeats the same tool call until the budget runs out.
Before and after
Same product instincts. New powers.
The training changes what you can do on your own.
You rely on no-code boxes or demos engineers dismiss.
After: You run live agents and see every step they take.
You wait two sprints for someone to change one retry rule.
After: You define agents, tools, MCP and memory yourself, in plain text.
You worry an unwatched agent will hallucinate in production.
After: You read the trace, name the failure, and add a judge or a human gate.
You're locked into one vendor or one framework.
After: You hand over plain-text specs any codebase can implement.
Ready for engineering
Your files become the spec
Test the workflow live. Then hand the exact RAC files to your engineers or AI coding tools, with no framework lock-in.
In numbers
The training at a glance
- minutes to your first live agent
- 0
- weeks, live and hands-on
- 0
- failure patterns you learn to fix
- 0
- in your browser, nothing to install
- 0%
Who it's for
Built for people who own the product
Product Managers
Turn an idea into a running agent without waiting for a sprint.
Senior PMs
Set the rules your team's agents follow, and check that they hold.
Technical PMs
Write agent specs engineers can build from on day one.
The outcome
What you leave with
Run live agents.Master Raw Agent Control.Become Agentsable
Questions
Quick answers
How much time does it take?
8 to 10 hours a week for 4 weeks: about 3 hours live, the rest on your own build.
Do I have to attend live?
Sessions are recorded, and the live Q&A recaps everything. Plan to attend at least 3 of 4 labs.
What do I leave with?
A working agent system, its failure profile, the checks around it, and the plain-text spec behind it.
When does it start?
Dates and tuition are announced once they're fixed. The waitlist hears first.
Become Agentsable.
Four weeks, live and hands-on, in your browser.
Dates and tuition are coming soon. Checkout is not open yet.