Interactive guides for product managers
Seven modules, written for PMs rather than engineers. Start with how agents work and the tools at a glance, go deeper on each tool, learn what to measure in production, then how to launch. Modules 1 to 5 and 7 follow the same example: a research assistant that takes a question, searches the web, reads the results and writes a short report. Module 6 takes those ideas into a regulated setting, a wealth assistant for investors. Every module has a hands-on demo you can click through, and quizzes to check yourself.
What an agent is, and how it gets from idea to launch.
The three tools at a glance: what each one does and how they fit together.
Builds on module 1. If agents are new to you, start there.
Deep dives
Each one goes further than module 2, with its own hands-on demo. They're best read after module 2.
Models, prompts, tools and MCP, retrieval, chains, memory and agents: the building blocks, and where they feel heavy.
State, branches, retries, checkpoints, human approval, time travel and testing: agents that pause, crash and recover.
Traces, cost and latency, prompt versions and evals: how to find out why an agent went wrong.
Module 6 · Running it in production
What to measure when an AI assistant talks to investors: 7 categories split into P0 launch blockers, P1 core health and P2 optimisation. Includes a spot-the-breach demo, how to prove a zero, wiring it up in Langfuse, an incident runbook and a P0 launch checklist.
Module 7 · Launch
The product work around launch: staged rollout with a kill switch, designing an experience people trust, all-in unit economics and pricing, build or buy, and the research assistant's first six weeks.
Just want to try things?
Switch between the LangChain, LangGraph and Langfuse demos without the lessons around them. Each one links back to its full module.
Reference
Every term from the seven modules in one searchable list, with links back to where it's explained, plus official sources to go deeper.
The first time a technical word appears, it has a dotted underline. Tap it for a one-sentence definition.
Each demo has failure modes you can trigger. Seeing how an agent goes wrong teaches more than watching it succeed.
Each module ends with a one-page summary and the questions to ask your engineering team. Module 1 also has a one-page agent spec template to download.