Practice AI skills

Every exercise gives instant feedback and explains why it is right — no account, no API key. Most exercises belong to a lesson; each one also works standalone from this page.

• Assemble a document-extraction prompt
  Intermediate · data extraction · verification — from “Extract data from documents”
• Assemble a spreadsheet-help prompt
  Intermediate · spreadsheets · verification — from “Spreadsheets: describe, don't guess”
• Assemble a structured-output prompt
  Intermediate · structured output — from “Structured output (JSON, simply)”
• Assemble a voice-matching email prompt
  Beginner · email · prompting — from “Email that sounds like you”
• Assemble an action-table prompt
  Beginner · meeting notes · structured output — from “Meeting notes to action items”
• Which request matches how an LLM actually works?
  Beginner · mental models · task fit — from “What an LLM actually does”
• Which prompt sets real boundaries?
  Beginner · constraints — from “Add constraints”
• Which prompt carries the context?
  Beginner · context — from “Give useful context”
• Which prompt shows instead of telling?
  Intermediate · examples · few-shot — from “Examples and few-shot”
• Which prompt aims at a result?
  Beginner · outcomes — from “Define the outcome”
• Choose the right coding task
  Beginner · coding · evaluation — from “What AI can (and can't) do with code”
• Choose the sound delegation
  Intermediate · judgment — from “When not to rely on AI”
• Choose the stronger follow-up
  Beginner · iteration — from “Iterate from the output”
• Choose the uncertainty-asking prompt
  Intermediate · prompting · verification — from “Ask for uncertainty”
• Which statement about tokens is accurate?
  Beginner · costs and limits — from “Tokens, simply”
• Choose when search changes the answer
  Beginner · research — from “Search vs generation”
• Which task fits a plain generative model?
  Beginner · foundations · task fit — from “What generative AI is”
• Compare two meeting summaries
  Beginner · summarizing — from “Summarize without losing meaning”
• Compare two replies and pick the one to ship
  Intermediate · evaluation · verification — from “Evaluate before you ship”
• Diagnose the layer
  Beginner · tool selection · mental models — from “Models vs products”
• Expect the variance
  Beginner · working with AI — from “Why answers vary”
• Fix the boss email prompt
  Beginner · clarity · instructions — from “Write clear instructions”
• Fix the forgetting chat
  Beginner · long input · context management — from “Context windows”
• Fix the triage prompt
  Intermediate · triage · structured output — from “Classify and triage text”
• Fix the unstructured summary
  Beginner · format — from “Specify the format”
• Inspect the suspicious citations
  Intermediate · verification — from “Inspect the citations”
• Order the long-document process
  Intermediate · long input — from “Work with long input”
• Pick the right tier
  Beginner · tool selection · costs and limits — from “Fast vs reasoning models”
• Plan your usage before you start
  Intermediate · judgment · privacy · verification — from “Using AI at work: judgment calls”
• Spot the fabrication
  Beginner · verification · hallucinations — from “Why AI makes things up (hallucinations)”
• Spot the injected instructions
  Intermediate · security · verification — from “Prompt injection, simply”
• Spot the unsupported claims
  Intermediate · verification — from “Find the unsupported claim”
• Spot the unsupported policy claims
  Intermediate · grounding · verification — from “Ground responses in your sources”
• Spot what should not be pasted
  Beginner · privacy · verification — from “What not to paste”
• Trace claims to their origin
  Intermediate · research · verification — from “Primary vs secondary sources”
• Vet the file-renaming script
  Beginner · coding · verification — from “Describe, vet, never blind-run”