Deepa Shenoy explains: Brutalist Workflow & Explainer Types

This is Deepa Shenoy's week three progress update with Humanitarians AI, narrated by Liam on her behalf. It summarizes what she learned from the Brutalist learning material and how she applied it in her own videos. The central idea is that the workflow connects planning, evidence, timing, quality control and human review, rather than being only a way to render a video.

A local toolkit, not an installer

The material introduced Brutalist as a local toolkit, not a traditional app installer. Its setup and doctor checks show which production dependencies are ready on your machine. A successful render still needs explicit human review, so a finished file is not the same as a finished video.

Beat sheets tie everything together

The beat sheet is what links narration, beats, visuals, timing and rendering. Narration sets the production clock, which means the visuals follow the spoken script rather than the other way around. Individual beat slots allow revisions without necessarily rebuilding the whole film, so a single beat can be changed on its own. The video also says visuals should support evidence, and it notes that she learned to compose landscape and portrait versions natively.

Three explainer types

The update distinguishes three kinds of explainer:

• AI Explainer. It teaches through explanatory visuals, not repetitive interface footage.
• CLI Explainer. It separates story structure from interface skin. The structure is problem, ask, code, output, revision, improved output, summary, handoff and outro.
• Deep Explainer. It links mechanisms across acts. It is not just a longer explainer, and its duration should follow what each idea needs.

Knowing which type fits a topic is part of the planning, since each one has a different shape.

Machines check, people decide

Automated quality control can check things like resolution and branding. People make the editorial decisions. The video treats these as two parts of one process rather than competing ones, with the automated checks handling what can be measured and the people handling judgment.

How she applied it

She applied the material in her own Humanitarians AI videos in a few concrete steps: planning beats, reviewing narration, making native formats, running quality control and watching the cuts. Her conclusion is that working through the material taught her not just how to render a video, but how planning, evidence, timing, quality control and human review work together in a film-as-code workflow.

Key takeaways

• Brutalist is a local toolkit, and its setup and doctor checks show which dependencies are ready.
• A beat sheet links narration, beats, visuals, timing and rendering.
• Narration sets the production clock.
• Beat slots allow revisions without necessarily rebuilding the whole film.
• AI, CLI and Deep Explainers each have a different structure, and a Deep Explainer is not just a longer one.
• Automated checks can cover resolution and branding, while people make editorial decisions.
• A successful render still needs explicit human review.

Try it yourself

Pick a topic you want to explain and decide which of the three explainer types fits it. Then sketch a beat sheet for it, with the narration for each beat and the visual that supports the evidence. Review the narration before building anything, and plan to watch the finished cut yourself.