Films
Each film is drawn frame by frame in the house system — the gradient descent actually runs, the fits are least-squares, the concept grid counts the real registry. No generated footage, no stock people, and nothing staged to resemble a real interface, a real record or a real person. Narration is synthetic and labelled as such below.
Plates
One plate per group, listing exactly what is inside it.
Provenance
Picture: drawn programmatically with Pillow from the app's own concept registry and from live arithmetic, then encoded with ffmpeg. Voice: synthetic text-to-speech (ElevenLabs). No image or video generation model was used, and no third-party course material is reproduced anywhere in this project.