The Loop Lab

Build the loop. Watch it run away. Engineer the stop.

"My students can chat with AI. I have no idea how to teach them to build one." If you run a CS classroom, a makerspace, a robotics team, or a hackathon, you know the gap: AI literacy for CS teachers has mostly meant slides and vocabulary, and there is no hands-on agent curriculum for the one skill the field now rewards, building and governing autonomous agentic loops.

The Loop Lab is that curriculum. Across a tested, term-length sequence of buildable labs, your students assemble an agentic loop from scratch, deliberately watch it drift off-goal and burn tokens, then engineer the guardrails that make it trustworthy. You teach the failure first and grade the judgment after.

Inside, teaching AI agents in the classroom becomes a lab you can actually run:

• The build-break-stop lab structures every session: surface a failure students can see, then teach the one guardrail that fixes it.
• A runaway demo that lands every time, so the class watches a live loop drift and burn before anyone touches a fix.
• A drop-in grading rubric for every lab, built on evals and verification, so you grade the judgment instead of the vocabulary.
• Project-based AI lessons in a proven sequence, field-tested across classroom and hackathon sessions before they reached this page.
• Agentic loop lesson plans ready Monday morning, from the first assembled loop to the engineered stop.

Read it and you will close the term with students who can build a loop, prove it works, make it stop, and explain when a loop is the wrong tool entirely, the durable judgment no model upgrade erases. This is one of the first hands-on agent curricula written for the classroom, by a working learner-builder who teaches the way the field actually advances: ship, break, verify, repeat.

For CS teachers, makerspace and robotics mentors, hackathon coaches, and instructors of advanced high-school and undergraduate courses who can teach building but have no curriculum for proving and stopping a loop. Part of the Build Agents You Can Trust series, in The Verifier's Library.