Practice queue
Exercises still to do
Only the terms whose “Try It Yourself” exercise you haven't marked complete — grouped by category, easiest first.
Loading your progress…
Agents (6 left)
- Agent1/5 · Easy
Try it: Pick one task you do every week. Write one sentence describing an AI agent whose only job is that task.
- Human in the Loop2/5 · Getting there
Try it: Pick one AI task in your business and decide who signs off on it before it goes out.
- Multi-Agent System3/5 · Trickier
Try it: Think of a team you know — a crew, a kitchen, an office. Name each person's role, then match each role to an AI agent.
- Nano AI-Driven Ecosystem3/5 · Trickier
Try it: List three small jobs in your week that could each be handled by a tiny specialist AI. What would happen if they shared information?
- Orchestrator3/5 · Trickier
Try it: Imagine you have five AI agents. Write down which one your orchestrator would call first when a new customer emails — and why.
- Autonomous Agent4/5 · Advanced
Try it: Write one rule you'd give an autonomous agent before letting it work alone. Example: 'Never spend over $100 without asking.'
Models (8 left)
- Inference1/5 · Easy
Try it: Open your weather app and look at tomorrow's forecast — you just watched an inference happen. Name two more inferences you saw today.
- Model2/5 · Getting there
Try it: Ask an AI assistant one question about your trade. Everything in its answer came from what its model learned.
- Token2/5 · Getting there
Try it: Paste a page of your own writing into an AI tool and ask how many tokens it is.
- Context Window3/5 · Trickier
Try it: In a long AI chat, ask it to repeat your first instruction. See whether it still remembers.
- Fine-Tuning3/5 · Trickier
Try it: List the documents you'd give an AI so it sounded like your business instead of a stranger.
- Multimodal3/5 · Trickier
Try it: Take one photo from today's work and ask an AI to describe what it sees.
- Neural Network3/5 · Trickier
Try it: Find one photo-based task in your work and ask whether a machine could spot the same thing.
- Temperature3/5 · Trickier
Try it: Ask the same question twice in a tool with a creativity setting — once low, once high.
Data (4 left)
- Training Data2/5 · Getting there
Try it: Name the three information sources you'd want an AI to learn from before it worked in your business.
- Embedding4/5 · Advanced
Try it: Think of two phrases your customers use for the same thing. That gap is what embeddings close.
- Retrieval-Augmented Generation (RAG)4/5 · Advanced
Try it: Pick one document your team gets asked about weekly. That's the first file you'd hand to a RAG system.
- Vector Database4/5 · Advanced
Try it: Write down three past projects. What details would you want a 'library organized by meaning' to remember so you could find similar jobs later?
Infrastructure (4 left)
- Automation1/5 · Easy
Try it: Write down the task you repeat most each week. That's your best first automation.
- Latency2/5 · Getting there
Try it: Time how long your most-used AI tool takes to answer. Anything over 5s hurts.
- API3/5 · Trickier
Try it: Name two tools you use that don't talk to each other. Connecting them would be an API job.
- Vector Database4/5 · Advanced
Try it: Name the one folder of documents you'd most like an AI to be able to answer from.
Fundamentals (8 left)
- Prompt1/5 · Easy
Try it: Write the same request two ways — vague and specific. Notice which one you'd rather hand to a new hire.
- Algorithm2/5 · Getting there
Try it: Write a four-step 'recipe' for something you do daily — like making coffee or returning a customer call. You just wrote an algorithm.
- Benchmark2/5 · Getting there
Try it: Write five real questions from your work and run them through two AI tools.
- Guardrails2/5 · Getting there
Try it: Write three things an AI should never do in your business without asking you first.
- Hallucination2/5 · Getting there
Try it: Ask an AI a very specific factual question about your industry, then check the answer against a real source.
- Machine Learning2/5 · Getting there
Try it: List one decision in your work that you make from experience. That's a candidate for machine learning.
- Prompt Engineering2/5 · Getting there
Try it: Take one prompt you used recently and add three details: who it's for, how long, and what tone.
- Calibration3/5 · Trickier
Try it: Compare an estimate (yours or an AI's) to the real result of one recent job. How far off was it? That gap is what calibration shrinks.