Mapping the code-as-video landscape --- deterministic to generative, with a learning-content lens
A map of the code-as-video field, from Manim and Remotion to Sora, sorted from deterministic to generative by what each tier offers learning content.
Longer writing, newest first. 39 in total, page 3 of 4.
A map of the code-as-video field, from Manim and Remotion to Sora, sorted from deterministic to generative by what each tier offers learning content.
An agent reads Akaunting's codebase and produces a 71-second walkthrough, which changes how Learning Program Owners, Designers, and Developers split the work.
A first experiment in treating video like code: scaffolding an Editframe project, iterating with an AI agent, and rendering a hello-world MP4.
A nine-level model of personal AI maturity, and why most rollouts stall at the cliff between using AI and building with it.
The decomposition approach that took a months-stalled training adaptation project to a stakeholder-validated proposal in seven days.
Claude Code's changes were on GitHub but not in my local files, because it works in a git worktree by default.
A workflow that takes an hour in Claude takes three to five in Copilot, and mapping Claude's skill system onto M365 shows the gap is architectural.
GitSync broke the git index every time I added a new note on Android, so I replaced it with Termux.
A custom web client that swaps the ADK camera feed for screen capture, so the agent watches the application instead of the user.
Arguing over translation versus localisation misses the point; thinking in layers of context maps cleanly onto a multi-agent architecture.