Use these notes to review a concept or read at your own pace. The interactive workspace above adds predictions, editable code, and actual run results.
IDEA 1
Plan a learning outcome and time to practice
A workshop plan connects an outcome with activities that help participants work toward it. This fictional library session teaches checking a claim against evidence. The plan must include a check-claim activity and a practice activity. That coverage rule is specific to this exercise; it does not claim alignment to a school curriculum standard.
Time includes setup, not just the interesting parts. Add the setup minutes and every activity's duration before comparing with the session limit. A 30-minute plan with 30 minutes of activities and five minutes of setup is actually 35 minutes. A facilitator should review whether the durations are realistic for the participants; passing arithmetic cannot establish teaching quality.
A complete 30-minute outline
- Setup. Distribute materials and explain the task.
- Demonstrate. Show how a claim connects to an evidence card.
- Practice. Let participants test a new fictional claim.
- Review. Confirm the full sum and both learning components.
The timing is a fictional planning estimate, not observed classroom performance.
Start the total before the activities
let minutes = input.setupMinutes;
for (const activity of activities) minutes += activity.minutes;Initializing with setup prevents an invisible time cost. The same idea applies to introductions, transitions, and cleanup when those are part of a project's specification.
Think it through: Activities take 12 and 18 minutes, plus 5 minutes of setup. Does the plan fit 30 minutes?
IDEA 2
Verify the supporting material and quantities
Each activity refers to source cards that support its intended content. The readSource tool returns a fixture review status and the activity IDs that the card supports. A card marked conflict, a missing card, or a card for another activity does not earn support merely because it has an ID.
Materials are another independent check: multiply sheetsPerStudent by the participant count for each activity and add the results. This project checks paper quantities only; it does not establish accessibility, room safety, or instructional effectiveness. Its best output is ready_for_review, because a facilitator still needs to examine the original activity, participants' needs, and logistics before running a real session.
Check the plan from three directions
- Source support. Verify the card supports this exact activity.
- Materials. Scale each activity's paper need by participant count.
- Coverage. Include both demonstration and original practice.
- Human review. Return the plan's checks and remaining decision to a facilitator.
Ready for review means the fixture checks passed. It does not book a workshop or approve teaching for real children.
A source ID alone is not evidence
source.status === "supported" &&
source.supportedActivities.includes(activity.id)The status and relationship must both be valid. This explicit fixture relationship is a transparent teaching simplification, not an automatic judgment of arbitrary real-world sources.
Think it through: A source card exists but is marked conflict. What should the planner do?
Put it into practice
Include setup in the time total while checking source support, paper quantities, and both required learning activities.
- Run the too-long plan and compare its activities with the setup time.
- Insert the correct initial time value.
- Run all cases, including the conflicting source card.
- Change the custom participant count and inspect how material requirements scale.
Your next experiment: Increase students to 11, then remove the practice activity. Which supply and learning-coverage checks change?
An educational plan needs supported content, time to practice, complete logistics, and a human review boundary.
Key terms
- Learning outcome
- An explicit capability or understanding a learning activity aims to develop.
- Coverage
- Whether a plan includes every required component of its specification.
- Facilitator
- The person who guides and adapts a group learning activity.
Sources and scope
Original Stemtiq teaching, reviewed 2026-09-14. The named researchers and organizations do not endorse this course. Classroom cases are authored exercises, not published findings.
Anthropic: Building effective agents
19 December 2024
Distinguish fixed workflows from model-directed actions; start simply, use tool observations as feedback, and set stopping conditions.
The article notes that its tooling landscape has changed. Its patterns do not establish that a more autonomous or complex system is better for every task.
Lewis et al. · NeurIPS: Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks
NeurIPS 2020
Combine information retrieval with generation so a model can use external passages while producing an answer.
The paper uses a trained neural retrieval and generation architecture. A keyword lookup illustrates retrieval but neither reproduces that architecture nor proves a retrieved claim true.
Anthropic: Demystifying evals for AI agents
9 January 2026
Define tasks, trials, and graders; inspect both execution records and final outcomes; repeat trials when model behavior varies.
A score depends on its cases and grading rules. Repeating a deterministic classroom case does not measure the variability of a live model.
OpenAI: Guardrails and human review
Reviewed 13 September 2026 · undated documentation
Distinguish automatic checks from approval decisions, pause sensitive tool requests, retain state, and resume after an application approves or rejects them.
Model-generated approval text is not authorization. Resume examples that automatically approve a request do not establish that a person reviewed it.