Lesson 1 / 3 min · Start here
From answering to acting
Distinguish a model, a workflow, and an agent by who chooses the next step.
Read, simulate, checkAI agent foundations · Grades 9–12
Understand the idea.
Watch it work.
Start with how a model generates an answer. Build toward an agent that uses tools, checks evidence, and knows when to stop. Each short lesson pairs an explanation with a simulation you can change.
Begin lesson 1 7 lessons · About 25 minutes · Free · No account needed
By the end, you’ll be able to…
- Tell a language model, a fixed workflow, and an agent apart.
- Give an agent a clear goal, useful tools, and the right context.
- Trace how a tool result changes the next action.
- Check evidence and test whether an outcome meets the goal.
- Set permissions and decide when the agent must stop for a person.
- Design and explain a small agent for a school task.
Lesson 2 / 3 min · The ingredients
Give the agent a useful brief
Separate what the agent should achieve, what it can use, and what it knows for this task.
Read, simulate, checkLesson 3 / 3 min · Inside the flow
Watch the loop change direction
Trace an action, its observation, and the next decision.
Read, simulate, checkLesson 4 / 3 min · Evidence before confidence
Make the answer earn your trust
Match a claim to a relevant source and spot what remains unknown.
Read, simulate, checkLesson 5 / 3 min · How you know it worked
Test the result, then improve it
Choose checks that measure the goal and use a failure to revise the design.
Read, simulate, checkLesson 6 / 3 min · Permission and human judgment
Decide what the agent may do
Separate information from instructions and put a human check before consequential actions.
Read, simulate, checkLesson 7 / 4 min · Bring it together
Design your first agent
Build a complete plan and explain how it can succeed, fail, or stop.
Read, simulate, checkA small vocabulary. Useful ideas.
- Language model
- A model that uses learned patterns to generate or process language.
- Token
- A unit of text a model processes, such as a word or part of a word.
- Agent
- Here, a system where a model can choose next steps and tools while pursuing a goal.
- Workflow
- A sequence whose path is arranged in advance by its developer.
- Tool
- A function an application lets an agent request, such as a search or calculation.
- Context
- The information available to the model for its current response or decision.
- State
- Task information an application keeps as a run progresses.
- Memory
- Stored information an application can retrieve for later use; not unlimited recall.
- Observation
- Information returned by an action, such as a tool result or an error.
- RAG
- Retrieval-augmented generation: using retrieved material when generating a response.
- Evaluation
- A test of an output or outcome against defined success criteria.
- Prompt injection
- An attempt to redirect an AI system through instructions in content it processes.
Then put the ideas to work.
Explore five field labs in coding, research, law, business, and engineering. Each has short teaching bites, working tools, and a fresh case to test your understanding.