Exports

Scientific papers are typically distributed as PDFs — convenient for humans, but terrible for machines, AI, and modern workflows.

ScienceStack transforms LaTeX source into three structured export formats that preserve the full semantic content of research papers:

• Markdown (.md) — Human-readable, works with Obsidian/Notion/VSCode, preserves all numbering
• JSON (.json) — Machine-native, optimized for LLMs and AI pipelines
• LaTeX (.tex) — Raw LaTeX with all macros expanded

All formats preserve equations, section numbers, cross-references, and document structure — making them superior to PDF extraction or generic converters.

---

The PDF Problem

PDFs flatten rich document structure into visual layouts, stripping away semantic meaning:

• Loss of structure — Sections, figures, theorems, and references are mashed into a page dump
• Broken math — Equations are often extracted incorrectly (superscripts and fractions collapse)
• No semantic cues — Citations appear as [12] instead of links to actual references
• Bad for AI — LLMs waste tokens on noise (line breaks, formatting artifacts)

---

How to Export

1. Navigate to any paper on ScienceStack
2. Click the Download button in the top-right navigation bar
3. Select your preferred format from the dropdown
4. Configure options (annotations, assets) and download

---

Markdown Export

Our Markdown export is purpose-built for research papers and significantly more robust than generic LaTeX→Markdown converters.

Key Features

• Complete numbering preservation — Sections, equations, figures, tables, and theorems all keep their original numbers
• Linkable cross-references — All \ref{...} commands become live markdown links
• Complete asset package (Pro) — Download with all figures and diagrams in optimized formats
• LLM-friendly annotations — Your notes are embedded as structured JSON in HTML comments
• Works everywhere — Compatible with Obsidian, Notion, VSCode, GitHub

---

JSON Export

Our JSON format is machine-native and optimized for AI applications, LLM ingestion, and programmatic analysis.

Why JSON Over PDFs for LLMs?

Key Properties

• Macros expanded — All \newcommand definitions resolved
• Stable IDs — Every block has a unique identifier
• Semantic types — Explicit tags for abstracts, proofs, definitions, etc.
• Resolved references — \ref{thm:main} links to the actual theorem block

---

LaTeX Export

Download the raw LaTeX source with all macros expanded and content in the correct order.

What You Get

• Macro expansion — All \newcommand, \def, and custom commands resolved
• Complete content — All \input and \include files merged in order
• Clean formatting — Unnecessary whitespace and comments removed
• Bibliography included — References appended as BibTeX entries