Academa: LLM-Generated Long-Form STEM Lecture Videos

**Academa proposes code‑generated lecture videos** Sina Atalay and Abdullah Geduk, both PhD students, have founded Academa, a startup that aims to transform how educational content is produced. The company’s core idea is to write lecture videos as code, which is then compiled into a video using computer graphics and text‑to‑speech (TTS) engines. This approach would allow large language models (LLMs) to author the content, while any errors could be corrected by editing the underlying code rather than re‑recording the entire lecture. The result is a continuously improving video that automatically incorporates fixes, ensuring that viewers always see the most up‑to‑date version.

**Continuous improvement and interactivity** Because the videos are stored as code in a repository, every review or correction becomes a versioned change that propagates to all future viewers. Unlike traditional recordings that lock in mistakes, Academa’s model supports iterative refinement. In addition, each video includes an AI‑driven chat interface that can answer questions about the lecture’s content, providing an interactive layer that mirrors the spoken material. The founders have outlined the technical details and potential use cases in a longer write‑up on their website, and they are open to answering questions from the community.