Wharton's Generative AI Labs released Primer, a free open-source tool that turns a teacher's lesson into a playable simulation for students, with nothing built until the teacher approves the plan.

Key Takeaways

  • Primer interviews a teacher about the lesson, the students, common mistakes and how success will be judged
  • It turns those answers into a design document the teacher can edit and must approve
  • A chain of AI agents then builds the simulation, and more agents guide students through it
  • The paper, prompts and code are free to use, but Wharton calls Primer an untested prototype

Wharton professor Ethan Mollick, whose lab built it, said Primer lets a teacher talk through a lesson and get back "a playable simulation for their students, grounded in pedagogy." The lab released the paper, prompts, templates and code under open-source and Creative Commons licenses.

The team says it has run classroom simulations for more than a decade, but they are rare because they are hard to build. AI can play characters, improvise and react to each student's choices, so teachers no longer have to script every path a student might take.

Also Read: McGraw Hill Puts School Lessons Into Claude For Teachers

How A Lesson Becomes A Simulation

An intake agent asks the teacher one question at a time about the lesson, the students' level, what they already know, common misconceptions, the teaching approach and how work will be graded. Teachers can speak or type and upload syllabi, lesson plans or rubrics.

The answers become a design document covering learning goals, scenario flow, characters, stakes and how the simulation adapts. Only after the teacher revises and approves it do the other agents get to work, and the finished simulation can be assigned to one student or a whole class.

Primer agent Its job
Intake Agent Interviews the teacher and records goals and materials
Design Agent Writes the design document for the teacher to review
Casting Director Turns the approved plan into mentor, practice and feedback characters
Image Generation Agent Creates the scene pictures
Experience Builder Puts it all together into a playable simulation

Not Ready For Every Classroom Yet

Wharton says AI models left to themselves help students too quickly, drift from the lesson goal and produce convincing roleplay without real teaching structure. Primer writes scaffolding, recall practice and feedback into each agent's instructions to stop that.

The lab calls Primer an experimental prototype, not a vetted classroom product. It has not been tested as a learning tool in real classrooms or reviewed against student privacy laws such as FERPA and COPPA or district rules.

Read Next: Teaching AI To Teach Reveals 20.5-Point Simulated Score Gain

Originally published on Fathom News.