Inspiration
Most teachers get very little practice before standing in front of a real class. A mirror doesn't raise its hand or look confused, and an observer sees you teach once or twice a term at best. We wanted a safe place to get real reps in, so we built a classroom of AI students you can teach without a real learner paying for your mistakes.
What it does
Roundtable is a practice classroom where you teach a live, voice-driven class of AI students.
- Pick a topic.
- Teach out loud. The students hear you, react with emotions, raise their hands, and ask questions.
- When class ends, get a debrief covering what worked, the confusion signals you missed, misconceptions you let slide, and concrete things to try next time.
How we built it
Roundtable is a web app with live speech transcription. Each AI student is a language-model character with their own personality, background, and mood. After class, the transcript is analyzed to produce feedback tied to specific moments. The app doesn't store your voice or screen, and the students are clearly labeled as fictional.
Challenges we ran into
- Making each student feel like a distinct person who is confused in believable ways
- Keeping speech, responses, and reactions fast enough to feel like a real conversation
- Keeping the debrief honest when a short class gives it very little to work with
What we learned
Good teaching feedback is about noticing: the unanswered question, the silent confusion, the misconception that got praised instead of corrected. Closed questions like "Does that make sense?" tell you almost nothing.
What's next
More subjects, custom student personas, and progress tracking across sessions so you can watch your teaching improve.
Built With
- api
- claude
- next.js
- node.js
- react
- typescript
- websockets