ml-contentlisted

Generate publication-ready ML content. **Manim is the default video pipeline**; matplotlib is fallback when the environment can't run Manim or the user insists. When we can render ourselves, we render → critique → re-render until the result is pixel-perfect. This skill compresses methodology developed across published ML carousels + videos + a full reverse-engineering of the github.com/3b1b/videos repo (Grant Sanderson's actual production code) and his explanation psychology. It treats ML content as applied research communication, not graphic design with science vocabulary on top. --- ## Workflow Decision Tree ``` User has... → Start at... ──────────────────────────────────────────────────────────── Topic / paper, no plan → Stage 0 (AHA) → full pipeline A 5-file recon bundle → Stage 1 (Audit) → Build → Critique A finished design-spec → Stage 3 (Build) → Critique A finished video / carousel → Stage 5 (Grounding pass) Edits on a posted piece → Grounding pass + correction comment ``` Always ask what they have and what they need. Don't assume full pipeline. --- ## The Five Locks (non-negotiable) Skip any one of these and the output devolves into AI slop. **1. AHA lock** — every project starts with a pre-built AHA moment. The single shift-in-perspective the viewer will leave with. If you can't articulate it in one sentence, **do not start the project**. Grant: