ml-content

What problem does it solve?

It turns dense ML research into publication-ready, audience-specific content (carousels, explainers, infographics, posters, and figures) while enforcing factual grounding and real 3D correctness.

Core Features & Use Cases

• Deep paper recon (5-file bundle): produces paper-summary.md, related-work.md, discussions.md, brainstorm.md, and a project README.md to lock understanding and reuse across posts.
• Audience-aligned hook + worldbuilder discipline: constructs a specific, high-leverage opening and a slide arc that matches the intended reaction map.
• Design locks for reliable visuals: applies a dark, phone-readable brand system with annotation rules and a differentiation constraint to avoid template slop.
• Locked render pipelines: generates static visuals via HTML→WeasyPrint→PDF→pdftoppm, 3D charts via matplotlib 3D, and motion explainers via Manim.
• Mandatory grounding pass: grades every checkable claim as CONFIRMED/WRONG/UNVERIFIED against primary sources before posting.

Quick Start

Use ml-content to generate an Instagram carousel for a specific ML paper by providing the paper topic or arXiv ID and asking for a 10-slide package with recon, a locked 3D audit, design spec, and a grounding-ready set of claims.