CreativeOperations

Teaching, cohorts, and “vibe coding” education

Support students and bootcamps with structured prompts, safe sandboxes, and reviewable artifacts—teachers stay in the loop.

What you build

Pedagogy-shaped assistance:

  • Scaffolded assignments: constraints, test harnesses, rubric alignment.
  • Office-hours style explanations that cite course materials—not random internet lore.
  • Integrity-friendly workflows: what is allowed collaboration vs solo work—encoded in prompts and environment.

Educators in shoutouts care that learners understand, not only ship.

Why CoWork OS is a strong fit

  • Sandboxes and approvals reduce “agent deleted prod” class accidents in classrooms.
  • Skills can embed syllabus rules and forbidden shortcuts.
  • Artifacts (repos, PDFs) make assessment possible—chat-only demos are weak evidence.

How to use

  1. Publish policy: what tools may run, what data may leave campus VPN.
  2. Template starter repos with tests students must pass.
  3. Require reflection notes alongside generated code—pedagogy check.
  4. Monitor for plagiarism patterns; agents make collusion weirder—stay explicit.
  5. Iterate prompts from where students actually get stuck.

Prerequisites

  • FERPA / local education privacy rules understood for your context.
  • Lab machines or cloud credits with hard spend caps.

Steps

  1. Pilot with TAs before students.
  2. Baseline without agents—know the honest difficulty curve.
  3. Layer hints level by level; hide full solutions until due.
  4. Collect failure modes; adjust scaffolding.
  5. Survey weekly: comprehension vs shortcutting.

Suggested prompts

  • “Explain why this works without giving the full solution.”
  • “Generate three failing tests the student should make pass.”
  • “List misconceptions common for week 3 on this topic.”

Launch readiness

  • Academic integrity statement updated for generative tools.
  • Support load estimated—agents increase question volume sometimes.
  • Accessible paths for students without newest laptops.

Common pitfalls

  • Solution leakage through over-helpful agents.
  • Unequal access if only some students have API keys.
  • Evaluating chatty output instead of runnable projects.
  • Burnout for TAs reviewing agent slop.