These notes are the ideas behind Agent School, Deployed's agent engineering program, written for someone who has never used a model, never called an API, and has no reason to feel behind. Each one answers a single question you might actually ask, in about ten minutes of reading. There are forty-two of them, grouped into eleven clusters, from what an AI model is doing when it answers you, to how you present a finished system to someone who will never read your code.
Two ways through
You can read them cover to cover. The clusters are shelves, grouped by topic, and the early shelves give you words the later ones lean on.
Or you can follow the path the program follows: as you move through the missions, your mentor skill points at each note when the problem it answers shows up in your own work. Hit a surprise bill in the warm-up and the note on tokens turns up. Build a pipeline that fails quietly in Project 2 and the note on failing loudly shows up. Reading a note right when you need it sticks far better than reading it in advance, and the path deliberately cuts across the shelves. It runs like this:
- Week 0: 0-1 · 1-1 · 1-2 · 3-1 · 3-2 · 3-3 · 3-4
- Warm-up: 6-1 · 8-1
- Project 1: 1-3 · 2-1 · 2-2 · 4-1 · 4-2 · 6-2 · 7-1 · 7-2
- Project 2: 4-3 · 4-4 · 6-4 · 6-5 · 7-4 · 8-2 · 9-2 · 10-1 · 10-2
- Project 3: 5-1 · 5-2 · 5-3 · 5-4 · 7-3
- Project 4: 2-3 · 6-3 · 7-5 · 9-1 · 9-3
- Capstone: 7-6 · 8-3 · 8-4 · 10-3 · 11-2 · 11-3
- Every project ends with: 11-1
Either way works. Skipping ahead and coming back works too.
What every note looks like
Every note follows the same five beats, so you always know where you are:
- A question as the title, one you might ask out loud.
- The idea in plain words, with exactly one analogy.
- A story that actually happened to us in production, including what it cost, because mistakes we admit are more useful than lessons we polish.
- A two-minute experiment you can run in any AI chat, with no setup and no keys.
- A short paragraph on what this means for the thing you are about to build.
Do the experiment. It is the cheapest way to turn "I read about it" into "I've seen it happen."
The trick that makes them yours
These notes live as plain files in your own course repo, next to your project. That means your agent, the AI assistant working alongside you, can read them too. Open any note and ask it to explain the idea again using your project as the example. Ask for a different analogy if mine did not land. Ask it to quiz you, or to argue the opposite side so you find the holes in your understanding.
A note is a starting point. The version that matters most is the one your agent rebuilds around the thing you are making.
Start with whichever question is nagging you most.