AI Course Experiments
Experiments exploring AI-assisted course creation, avatars, automated content production, and educational workflow design.
Example Output
One of the AI-generated lesson videos produced during the project.
Published Output
The project ultimately resulted in a published introductory course on moxibustion through Flow Temple.
Overview
This project began as an attempt to better understand and organize knowledge around Traditional Chinese Medicine and moxibustion.
While acupuncture is widely known, moxibustion remains relatively unfamiliar despite being part of the same tradition. To deepen my own understanding, I started collecting books, articles, research papers, and educational resources.
What began as a personal learning project gradually evolved into an experiment in AI-assisted course creation and knowledge synthesis.
Why I Started
I often learn best by building.
Rather than simply reading about a topic, I like creating structures around it: notes, frameworks, visualizations, and educational material.
For this project, I wanted to explore whether modern AI tools could help transform large collections of source material into a coherent learning experience while simultaneously helping me deepen my own understanding of the subject.
Core Question
Can modern AI tools help transform large amounts of source material into structured educational content without losing the quality of the underlying knowledge?
Workflow Diagram
Key Experiments
Knowledge Synthesis
Before building the course, I spent considerable time assembling and organizing source material. NotebookLM became a useful companion for exploring themes, identifying knowledge gaps, and surfacing connections across different sources.
- Books on TCM and moxibustion
- Research papers and online articles
- Practitioner resources
- Topic exploration & curriculum design
AI Slide Generation
Generating course decks from structured outlines.
- Visual learning materials
- Rapid iteration
- Utilizing Gamma
AI Avatar Production
Testing synthetic instructors using HeyGen.
- Voice cloning
- Pronunciation handling
- Gesture controls
Automated Production Workflows
- End-to-end content pipeline
- Reduced manual effort
- Repeatable process
Technology Stack
- Knowledge Synthesis: NotebookLM
- Content Generation: LLMs (Gemini, Claude, GPT-4)
- Slide Generation: Gamma
- Video Avatars: HeyGen
- Automation: Custom pipelines
Lessons Learned
The most interesting discovery was not that AI can generate content. That was expected.
The more interesting realization was how quickly the bottleneck shifted. Creating slides, narration, videos, and learning assets is becoming increasingly automated.
The challenge is no longer production.
The challenge is deciding what is worth teaching.