Coming October 2026

LaTeX produces beautiful documents.
It also makes you wait.

texlode eliminates that wait. You see your typeset output as you type — every line break, every equation, every margin adjustment — with the full typographic quality of real LaTeX.

Get Notified See the Difference
chapter-03.tex
\section{Quantum Entanglement}
When two particles become entangled,
measuring one instantly determines the
state of the other, regardless of the
distance separating them. This
\begin{equation}
|\Psi\rangle = \frac{1}{\sqrt{2}}
(|01\rangle + |10\rangle)
\end{equation}
Compiled in 0.5ms LuaTeX

The gap between thinking and seeing

When you type in a word processor, you see the result immediately. Visual feedback is part of the thinking process. When you type in LaTeX, you see source code. The rendered result arrives seconds later. The feedback loop is broken.

Overleaf / VS Code

  • 2 to 20 seconds per compile
  • Full document recompilation
  • Slower on larger documents
  • Full LaTeX compatibility

Typst

  • 100 to 200 ms compilation
  • New markup language
  • No CTAN package ecosystem
  • No microtypography

texlode

  • Under 1 ms per paragraph
  • Standard LaTeX source
  • Real LuaTeX engine
  • Knuth-Plass line breaking

What changes when feedback is instant

You stop guessing and start seeing. Adjust a margin, refine a paragraph, tweak an equation — and the typeset result updates before your finger leaves the key. Layout decisions that used to take compile-wait-check cycles now happen in flow.

Better layout decisions

See how your paragraph actually breaks across lines as you write it. No more compiling to check if a sentence wraps awkwardly or an equation overflows the margin.

Same speed at 500 pages

Editing a paragraph in a 500-page textbook feels exactly like editing a one-page letter. Document size does not affect your typing experience.

Your existing documents work

texlode runs the real LuaTeX engine. Your .tex files, bibliographies, custom class files, and packages from CTAN's 6,000+ ecosystem just work. Nothing to port or rewrite.

Real typographic quality

Knuth-Plass line breaking, proper hyphenation, optical margin alignment. The same algorithms that make LaTeX output beautiful — now visible while you write, not after you compile.

Performance that scales

The fast path is O(1) with respect to document size. Whether you are editing a single-page letter or a 500-page textbook, paragraph compilation time stays the same.

0.5ms
Paragraph compile time
<7ms
End-to-end latency
100×
Message size reduction
26×
Cache speedup

Where texlode fits

Every editor makes different trade-offs between latency, typographic quality, compatibility, and collaboration.

System Latency At 300 pages Typography Compatibility Collaboration
Word / Google Docs <100 ms <100 ms Low N/A Yes
Typst 100-200 ms 200-500 ms Medium No Planned
Overleaf 2-5 s 10-30 s High Full Yes
VS Code + LaTeX Workshop 5-20 s 20-60 s High Full No
texlode <1 ms* <1 ms* High Partial Git**

* Per-paragraph fast path. Full document compile for page positions runs in the background.
** Git-based collaboration at launch. Server-based real-time collaboration planned.

Who is texlode for?

texlode is designed for writers who value both typographic quality and responsive editing.

Technical Authors

Writers of textbooks, dissertations, and research papers who use LaTeX for its typographic quality but want to see results as they type.

Book Authors

Authors working on long documents where traditional LaTeX compilation times scale with document size. texlode does not.

Developers

Engineers building typesetting systems, editor tooling, or integrations who want to understand how to make TeX responsive.

The architecture, documented

texlode: Architecture of a Real-Time LaTeX Editor documents the design decisions, trade-offs, and open problems involved in making TeX responsive enough for interactive editing.

Part I

The Problem

The 100-millisecond barrier, a survey of existing LaTeX editors, and why TeX's algorithms are worth preserving.

Part II

The Architecture

The persistent daemon, fast and slow path separation, the three-level cache, and canvas rendering via display lists.

Part III

Deep Dives

LuaTeX internals, the IPC protocol, handling floats, footnotes, cross-references, performance engineering, and frontend architecture.

Part IV

Reflections

Architecture trade-offs, dead ends and failures, design principles, and the future of real-time typesetting.

Get notified at launch

Sign up for updates on the editor release. No spam, just milestones.