Learn
Agent University
Lessons coming soon. Here is the proposed outline, beginner first.
Agent University will be a set of short lessons on working with AI agents: what they are, how to give them work, and how to check what they did. Lessons are coming soon Planned. Nothing is published yet.
Who it is for
Anyone who wants to get real work done with an agent, starting from zero. The first lessons assume no technical background. Later lessons go further, for people who want to run experiments or connect their own tools.
Proposed lessons
This outline is a proposal and may change before the first lesson is published.
Beginner
- What an agent is. The difference between a chatbot, an agent and the AI model underneath. Why an agent has a name, a job and a record.
- Giving an agent a task. How to describe what you want, what "done" looks like, and what the agent should not touch.
- Approvals. Which actions wait for you, how to read an approval, and when to say no.
- Checking the work. Reading a receipt, telling a real result from a confident summary, and asking for the original.
Getting comfortable
- Notebooks and memory. What an agent remembers, where it is kept, and the difference between a note, a suggestion and an instruction.
- Your sources. Bringing in files, links and notes, keeping originals, and reusing one source in many notebooks.
- Privacy in practice. What an action sends and to whom, keeping secrets out of chats, and choosing whether to help with training.
Going further
- Running a fair comparison. Using Labs and Tilly to compare two models or two prompts, and reading the result honestly.
- Connecting tools. Adding the Guth MCP server to an AI client and using instruments and field guides safely.
- Improving an agent. How personal training is meant to work: examples, consent, testing on fresh examples, and switching on only what helps.
Until the lessons arrive
- Getting started covers the basics.
- Agents and AgentNotebook explain the main ideas.
- The glossary defines the words in a sentence or two.