Show HN: Academa – Long-form STEM lecture videos generated by LLMs

Online STEM education today relies on recorded lecture videos—Khan Academy, Udemy, Coursera, MIT OpenCourseWare, YouTube. A professor at a blackboard performs a presentation: writing equations while explaining, circling terms, drawing plots, and pointing to key ideas. Recording that yields a lecture video, but producing a good one is hard, and fixing a mistake after release is nearly impossible; you either ship a flawed video or redo the entire production.

The post proposes a paradigm shift: write lecture videos as source code. Instead of recording raw video, describe the teacher's actions in a structured language, e.g., 'say "Look at this square." while: draw square' and 'write "A = s^2" while explaining the area formula.' A hypothetical compiler turns this code into a video using text-to-speech and computer graphics. Editing becomes as simple as changing the source and recompiling, just like software.

This code-based representation unlocks AI capabilities. Because LLMs work on text, once lectures become code, LLMs can create lectures for obscure theorems, niche engineering methods, or research papers—subjects with too little audience to justify human production. They can also translate a lecture into 80+ languages, producing a first-class version rather than a dub or subtitle. Finally, code-based lectures enable an AI-native student experience: pausing to ask a question and getting a real-time generated video answer, or having a course personalized to individual learners.