LearnAILevel 1.5 · Coach HQ ← Coach home

For coaches & teachers

Level 1.5 · Coach HQ

Everything to teach the AI-Agents track: three ready-to-present decks with coach notes, three Google Colab notebooks where students build real agents, and a printable energizer for every session.

🧭 Where this fits

Level 1.5 turns the Agents roadmap (10 deep dives) into a taught workshop. It bridges Level 1 · Foundations and Level 2 · Build agents (Weeks 2–5). See all levels on the coach home.

The track at a glance

Before you start

  1. Free Gemini key for you and each student — aistudio.google.com/apikey (free, no card).
  2. Present the deck: open a session deck and press S for Speaker View (coach notes + timer + next slide). F = fullscreen.
  3. Open the notebook: download the session's .ipynb, then in ColabFile → Upload notebook.
  4. Print the energizer and grab any props it lists.
🖨️ Printable run-sheets

One-page cue sheet per session — running clock, prep checklist, block-by-block cues: open the Facilitator Run-Sheets.

Day-one icebreaker (foundational)

Human Calculator → Human LLM — students act as a calculator (same answer every time), then as an LLM (a different plausible answer each time). The fastest way to make "AI predicts likely words, it doesn't compute one truth" click. Great before Session 1.

The three sessions

Session 1 · Stations 1–3

Meet the Agent

What is an agent · the LLM engine · the agent loop → build one

Session 2 · Stations 4–6

Powering the Agent

Tools & actions · prompting · memory (embeddings)

Session 3 · Stations 7–10

Real-World Agents

Architectures · building · testing · safety (prompt injection)

Hands-on labs · build an agent with API + MCP calls

Two deeper build labs for Stations 7 & 8 — the same agent (a REST API call + an MCP call to solve a problem), one visual, one in code. Great as a stretch, a bonus session, or a bridge to Level 2.

Station 7 · Architectures · visual

Build an agent visually (Langflow)

Drag-and-drop · Agent + API Request + MCP Tools · zero code

Station 8 · Building Agents · code

Build an agent in code (Google ADK)

Python · ADK agent + REST API tool + MCP toolset · free on Gemini

Tip · Running the room

Get to the build fast

The ideas are the setup; the agent they build is the payoff. Keep theory tight, run the energizer to reset energy, then give the longest possible stretch in Colab. Pair students up so no one is stuck on a setup or tool error.


Self-study students can read the Level 1.5 roadmap & deep dives first, then come build. Part of learn.millionroots.com.