md2rich

2026-09-07 · 6 min read

MarkFlowy 2026: Markdown Editor Rebuilt for Large Documents

A Tauri + ProseMirror desktop editor that opens a 2 MB Markdown file in about a second, ships with built-in DeepSeek / OpenAI / Ollama, and fits the authoring end of a write-once-publish-anywhere stack.

I keep a folder called drafts/ on my laptop that holds long Markdown files — research notes, half-finished posts, a slowly growing book chapter. The folder is the bottleneck of my writing week. When the editor slows to a crawl at 500 KB, I stop adding words. When it crashes at 2 MB, I start splitting the file, and once I split, the thread is gone. So when MarkFlowy hit Show HN on September 6, 2026 with the line "A 2 MB Markdown file opened in around 1 second in testing," I went looking for the catch. There isn't one, and this article is why that one-second figure matters for the rest of your Markdown stack.

What v0.100.1 actually shipped on September 6

MarkFlowy is an independent developer's project — repository drl990114/MarkFlowy, AGPL-3.0, 2,382 stars, 93 forks, 23 watchers as of 2026-09-07. The first Show HN was 2025-08-28; v0.100.1 was tagged and released on 2026-09-06 at 10:16 UTC. The release notes are small and concrete, which is itself a signal: file-tree "Copy to…" menu, full-width toggle that no longer requires reopening the document, restored "Type '/' for commands" placeholder in WYSIWYG mode, list Tab / Shift+Tab nesting that preserves task state and undo history, and automatic Markdown formatting while typing tables, task lists, callouts, and inline math. The big-number change underneath those notes is the editor rebuild — the part that makes the 2 MB-in-1-second figure real.

Why "fast on a large file" is a real engineering claim

Open any electron-era Markdown editor with a 2 MB file and you can watch the page repaint lag as the renderer tries to lay out the whole document. MarkFlowy picks three tools that don't have that problem:

Distribution and the macOS Gatekeeper wrinkle

MarkFlowy ships for Windows, macOS, and Linux. macOS has two builds — Apple silicon (aarch64, 20.4 MB) and Intel (x64) — and the team does not have an Apple Developer ID, so the first launch hits Gatekeeper's "cannot be opened because the developer cannot be verified" wall. The README workaround is the standard unsigned-app incantation: open Terminal, cd /Applications, run xattr -cr MarkFlowy.app, reopen. Windows ships an online installer, an offline installer (266.5 MB with the WebView2 runtime bundled), and a portable zip. Linux ships a .deb, an .rpm, and an AppImage. A Flatpak is listed on FlatPark, but the page is not yet live — which matters for Linux users who want auto-update and sandboxed install, because Flatpak is the only path that gives both for a Tauri app today.

The AI features are real, not a wrapper

A lot of "AI Markdown editor" entries in 2026 are a chat sidebar bolted onto a textarea. MarkFlowy's AI features land closer to the document: built-in Copilot, one-click export of a chat dialogue into the document, translation to any language on the current selection, and article summaries. Out of the box, it integrates with DeepSeek, OpenAI, and Ollama, which means the same UI works with a hosted endpoint, a BYOK key, or a local model running on the same machine. For a writer on a long draft, that last option is the meaningful one — you can summarize chapter six or rewrite a paragraph through a local Llama or Qwen without the text leaving the laptop, then convert the cleaned Markdown to rich text with md2rich for publishing.

Where MarkFlowy fits, and where it doesn't

MarkFlowy is an authoring tool, not a publishing tool. The output it produces is a .md file on disk, optionally synced through Git. Once the draft is clean, the Markdown still has to be converted into rich text for the platforms most readers actually scroll: LinkedIn's article editor, X Articles, Medium, Notion, a Confluence page, a Substack post. That conversion is the part md2rich exists for — paste the Markdown, copy the rendered HTML, drop it into the destination editor. The two tools sit at opposite ends of the same loop: MarkFlowy is where you finish the draft, md2rich is where you finish the post.

For comparison, the other Markdown editors that have shipped to this audience in the past month cover adjacent slots: Markdown Buddy is the small native macOS slice (4.8 MB sandboxed AppKit app with a Quick Look extension), the headless / block / WYSIWYG framework tier (CodeMirror 6, Milkdown, BlockNote, Tiptap, Lexical, Plate) is the framework-level slot for teams building their own editor, and MarkFlowy is the long-document desktop editor where you actually write the book.

A short Markdown-to-rich-text sample

Here is a Markdown block as it would look in MarkFlowy's source mode, and how the same block renders after a single paste through md2rich.

## What I learned shipping MarkFlowy

The rebuild cut our open-time benchmark from **6.2s to 0.9s** on the same
2 MB file. The biggest win was rendering only the viewport — the
other 90% of the document is just buffered bytes.

1. Pick a runtime that fits your worst-case file, not your median.
2. Use ProseMirror (or its derivatives) when documents exceed ~200k chars.
3. Ship a Tauri shell, not Electron, if cold-start matters.

Try md2rich to convert this to rich text for LinkedIn.

The same block, after a single paste through md2rich, becomes a styled <h2>, three list items with working checkmarks, the inline <strong> emphasis, and a clickable anchor. Nothing leaves the browser — md2rich runs entirely client-side — so the post lands on LinkedIn looking the way you wrote it, not the way a third-party API chose to render it.

Should you switch?

If your Markdown files stay under a few hundred kilobytes, Typora, Obsidian, iA Writer, and the rest of the editor tier are still faster to learn. The case for MarkFlowy starts the moment your draft crosses ~1 MB and you notice the lag. At that point you have three options: switch editors (MarkFlowy), split files (loses the thread), or give up (loses the post). The first option is the only one that preserves both the document and the writing habit.

FAQ

What is MarkFlowy and why did it get a Show HN on September 6 2026?

MarkFlowy is an open-source cross-platform Markdown editor by developer drl990114 (github.com/drl990114/MarkFlowy, 2,382 stars as of 2026-09-07, AGPL-3.0). It first hit Show HN on 2025-08-28, and a v0.100.1 rebuild on 2026-09-06 (Show HN story 49586334) reopened the conversation specifically because the editor now opens a 2 MB Markdown file in about one second. That is the verifiable claim on the README: "A 2 MB Markdown file opened in around 1 second in testing." For anyone who writes a book-length Markdown draft, a long-form research report, or a multi-year personal knowledge base, that one-second figure is what separates an editor you keep using from an editor you abandon by chapter three.

How does MarkFlowy reach that one-second open time on a 2 MB Markdown file?

Three engineering choices. First, the runtime is Tauri, not Electron, so the shell is written in Rust and the binary is under 25 MB on macOS (MarkFlowy_v0.100.1_aarch64.dmg is 20.4 MB; the AppImage is 98.1 MB because it bundles GTK, but the macOS app itself stays lean). Second, the editing core is ProseMirror, the same engine that powers Tiptap and Atlassian's editor, which renders visible content only and streams the rest of the document on demand. Third, the source says the team rebuilt the editor specifically to support documents with over 200,000 characters, which in Chinese / English mixed writing lands around 2 MB once images are excluded.

What platforms does MarkFlowy support and how is it distributed?

Windows, macOS, and Linux. macOS ships separate Apple silicon (MarkFlowy_v0.100.1_aarch64.dmg, 20.4 MB) and Intel (MarkFlowy_v0.100.1_x64.dmg) builds. Because the macOS build is not notarized by an Apple Developer ID, Gatekeeper blocks the first launch with "cannot be opened because the developer cannot be verified"; the README workaround is xattr -cr MarkFlowy.app inside /Applications, then reopen. Windows ships an online installer, an offline installer that bundles the WebView2 runtime (266.5 MB), and a portable zip. Linux ships a deb, an rpm, and an AppImage. A Flatpak is listed on FlatPark but the page is not yet live.

Does MarkFlowy have built-in AI features?

Yes. The README lists Copilot, one-click export of dialogues, translation of articles to any language, and article summaries as built-in features. Out of the box, MarkFlowy integrates with DeepSeek, OpenAI, and Ollama, so you can point it at any OpenAI-compatible endpoint, including a local Ollama model, without a proxy. AI actions run inside the editor on the selection or on the whole document, and the rewritten text replaces the original block in place — which is exactly the workflow MRDown-style diff editors are designed to review after the rewrite lands.

Where does MarkFlowy fit in a write-once-publish-anywhere Markdown stack?

MarkFlowy owns the authoring end of the pipeline: long-form drafts, book chapters, research reports, and knowledge-base editing on a native desktop where 2 MB files stay snappy. Once the draft is clean, the Markdown source still has to be converted into rich text for LinkedIn, X Articles, Medium, Notion, or a Confluence page. That publish step is where md2rich fits — paste the cleaned-up Markdown into md2rich.com, copy the rendered HTML, and drop it into the platform's editor. MarkFlowy is not a publishing tool; it is the editor you finish the draft in, and md2rich is the tool you finish the post in.

Try it

If you write long Markdown and your current editor slows down past chapter two, give MarkFlowy a shot. Then, when the draft is done, paste it into md2rich to turn it into rich text for LinkedIn, X Articles, Medium, or Notion. Two tools, one Markdown source, no copy-paste cleanup.