Skip to content
Trace

Learn how AI works.
Build with it.

Readings, interactive labs, and agent projects for high school. Students build an agent that runs. Teachers assign what fits and can see every step of the work.

Try a lesson

Maya

Grade 8

Vocab coach

Quizzes you on the week's Spanish words and remembers what you've seen.

Build one of your own

Read it. Try it. Build it.

A lesson explains one idea. A lab lets you change it and watch what happens. A project makes you use it. All three below are the real thing, so go ahead and touch them.

Step 1

Read a lesson

Short articles in order, with a check-yourself question after each idea. Answer it below. Trace keeps your place in the chapter.

My learning
Language models

Grade 10 · Sampling

Where the randomness comes from

Ask about this lessonIn progress

At every step a model has a spread of likely next tokens, not one. Temperature decides how far it will wander from the safest choice. Turn it down and the model repeats itself. Turn it up and it starts to surprise you.

Next token after The night was
  1. cold62%
  2. freezing21%
  3. quiet9%
  4. long5%
  5. purple3%

The lab on the next page lets you set it yourself and watch three answers to the same question drift apart.

Question 3 of 5
You set temperature to 0 and ask the same question twice. What comes back?
My learning
Data science

Grade 10 · Fitting a line

In progress

Line of best fit

Goal: Move one point until R² drops below 0.8.

00224466881010Hours studiedQuiz score

Drag this one down

The line it learned

y = 0.79x + 1.10

Total error0.85
How well it fits (R²)0.971
xyerror
1.02.2+0.31
2.32.6−0.32
3.54.1+0.23
4.84.4−0.50
6.06.2+0.35
7.46.6−0.36
8.78.3+0.31
Not yet. The dots follow the line closely, so R² starts near 1.
Step 2

Play with the idea

Drag a dot and the line of best fit moves with it. The equation and the error update as you go, so a model learning from data stops being abstract.

Step 3

Build an agent

Connect a model, two tools, and a reply, then run it. This one runs on a real model right here, and you can follow every step it takes.

Game helper

The canvas is empty.

Building…

Can I grow melons before winter?

Waiting for the build to finish.

Message your agent

Six topics, from data to agents

Each one starts with a reading, tests the idea in a lab, and ends with something you built. Start with any of them.

Data science

Explore data, spot patterns, and use statistics to make sense of them.

Explore

Machine learning

Train a model on examples, test its predictions, and find where it gets things wrong.

Explore

Language models

See how a language model generates text, and how prompts and settings change its answers.

Explore

AI agents

Give a model tools and memory, then watch it work toward a goal step by step.

Explore

Applying AI

Use AI on a real problem, then check whether its answer holds up.

Explore

Building AI applications

Build, test, and ship an AI feature the way an engineer does.

Explore
Three students of different ages learning together with a laptop

Start with no coding experience

The first lessons use examples students already recognize. Code and model settings arrive later, when a lesson needs them.

  • No coding or prior AI coursework required.
  • Move from everyday examples to prompts, tools, and agents.
  • Each project keeps its earlier runs, so a revision has context.
Trace for students

Everything is written. You choose what to assign.

Pick readings, labs, quizzes, and projects for your class. Open a submission to read the student's answers and follow every step the agent took.

  • Readings, labs, quizzes, and agent projects, ready to assign.
  • Classroom AI credits come with your plan. Students never sign up for or pay a model provider.
  • Every submission keeps the project version and every run behind it.
Trace for educators
Assignments
3 of 3

Language models

3

Build a research agent

18 of 26 students complete · 3 waiting for review

Maya R. · Submitted agent

Why did you add a second tool?

My first run made up a source, so I gave it search and told it to quote the page it found.

  1. model turnResearcher → Model

    Asked for the Search tool

  2. tool callResearcher → Search

    search("tide pool species California") · 4 results

  3. tool callResearcher → Search

    Timed out after 8s. Maya changed the query and reran.

  4. message sentReply

    Quoted two sources.

Reviewed records that you looked. It is not a grade.

An example class. Names and work are made up, and Trace does not show invented scores. Pick an assignment to open a submission.

Set it up once for the whole school.

Teachers get material to assign. The school manages sign-in, access, and billing in one place.

Trace for schools
  • Single sign-on and rostering, with roles and access managed in Trace.
  • Classroom AI credits included in the plan. No student accounts with a model provider.
  • One bill for the school, and a rollout plan for the teachers who start.

FAQ

The things teachers and students ask before they start.

No. Students build agents by connecting blocks on a canvas, and each reading explains the idea before any code appears. Some advanced lessons include short Python examples.

Build your first AI agent.

Read a lesson, test the idea in a lab, then connect the blocks and run it.

A free account starts with 2,000 credits a month for running your agents.

Create a free account