Use the AI's copy button
ChatGPT, Claude, and Gemini all expose a copy button that preserves the underlying markdown — code fences, lists, and LaTeX included. Manual text selection loses all of it.
Practical writeups on the conversion workflows we ship for. Skim, bookmark, ship.
Copy from ChatGPT, paste, export DOCX. Headings, lists, code, and equations all carry over.
5 min readWorkflowAnthropic's Claude (Sonnet, Opus, Haiku) emits clean GFM markdown — perfect input for our DOCX exporter.
5 min readMathInline and block formulas, matrices, Greek letters — and the recent fix that keeps PDF fraction bars in the right place.
8 min readReferenceA4, smart page breaks, KaTeX via SVG foreignObject, syntax-highlighted code, CORS-inlined images. What works, what doesn't.
7 min readProductivityDrop up to 20 markdown files at once. Concurrency, ordering, and the gotchas you only learn after the third try.
6 min readConversionGet clean output from messy HTML — code blocks, tables, links, and lists translated faithfully.
7 min readChatGPT, Claude, and Gemini all expose a copy button that preserves the underlying markdown — code fences, lists, and LaTeX included. Manual text selection loses all of it.
Inline `$x$` and block `$$ ... $$` render via KaTeX in HTML/PDF and translate to native Office Math (OMML) in DOCX. Equations stay editable in Word.
Drag a folder of markdown files onto /batch. Three convert in parallel, the rest queue, everything bundles into one ZIP.
PDF, HTML, Markdown, and the live preview run entirely in your browser. DOCX export POSTs the markdown to our Pandoc route for higher-fidelity Word output — temp files only, deleted immediately, no logging of content. Set-it-and-forget-it for sensitive docs: the local-only DOCX fallback always works too.
Your current draft is mirrored to localStorage and restored on next visit. The History panel keeps the last few conversions for quick re-export.
When the server-side Pandoc route is reachable, exports go through Pandoc for highest fidelity. Otherwise the client-side `docx` package handles it. Both are good — Pandoc just wins on tricky edge cases.