Editorial: LaTeX tooling rarely gets flashy coverage, but small correctness tools can dramatically change workflows for researchers and technical authors. Today’s pick is ExactTeX — a pragmatic attempt to give LaTeX the kind of name-safety and editor feedback we've come to expect from programming languages, without forcing a rewrite.
In Brief
ExactTeX: a gradually-typed superset of LaTeX
Why this matters now: ExactTeX brings incremental, name-aware checks and editor diagnostics to LaTeX projects, letting authors catch broken labels, misplaced references, and unsafe renames before compiling PDFs.
ExactTeX is a drop-in superset: rename .tex to .xtex and your files keep working byte-for-byte, but you can selectively annotate sections, figures, citations, and other named entities to get checks and safe refactors. The project ships a zero-dependency compiler plus three identical surfaces — a CLI, an LSP for editor integration, and a WebAssembly build that runs inside a browser editor called Vitela — so diagnostics are consistent whether you run locally or in the cloud. According to the repo, the design is intentionally “gradual”: unannotated LaTeX is left untouched so partial adoption never generates spurious failures. (See the ExactTeX repo.)
Why this feels different (short)
ExactTeX focuses on the kinds of correctness checks authors actually miss: named labels, cross-references, and safe renames inside long documents. It also reports a coverage metric (how much of the file is under contract) and embeds a change/revision model inside the file that aims to mirror tracked changes in Word — useful for long collaborative papers where tracking edits and name changes gets messy. Early feedback is still thin, but the implementation surfaces and core guarantees look pragmatic enough to try on real projects.
Deep Dive
ExactTeX: A pragmatic type system for documents
Why this matters now: ExactTeX’s gradual typing and identical CLI/LSP/WASM surfaces make it immediately usable for authors and editors, lowering the barrier for safer document editing across local and cloud environments.
The core idea behind ExactTeX is simple and worth repeating: give authors a way to declare intent around named entities in their document and get machine-checkable guarantees without forcing an all-or-nothing migration. The project author puts it plainly:
"the type system is gradual, and that is the whole design."
That phrase matters because it defines both the product and the risk model. Gradual typing, in programming languages, lets you add static checks incrementally — critical when working with large, messy codebases. ExactTeX applies that same philosophy to documents: you can start by annotating a few figures or sections, get meaningful diagnostics and refactor support for those parts, and leave the rest of your LaTeX alone. Because unannotated code is treated as "unknown" rather than "incorrect," you avoid the common upgrade pain of false positives that kill adoption.
Practically, ExactTeX provides three identical surfaces: a CLI compiler, an LSP server for editor features (hover, completion, diagnostics), and a WebAssembly build powering a browser editor named Vitela. That parity is not just convenient — it's a product-level decision that reduces surprise. If your editor shows a label as missing, the same check will flag it in CI and the same diagnostic appears in the browser build. This consistency is the kind of thing that makes adoption realistic for collaborators who live in different environments: Overleaf users, local Emacs/Vim/VS Code folks, and reviewers working in the browser.
There are obvious limitations and open work. The change/revision model and the UX around in-file tracked edits are still rough and will need real-world testing on collaborative workflows (journals, conferences, thesis committees). Transport guarantees between the three surfaces are reportedly solid at the moment, but integration with popular hosted LaTeX platforms (Overleaf, ShareLaTeX forks) and journal submission toolchains will determine whether ExactTeX stays a niche power-user tool or becomes part of standard academic workflows. Still, for anyone managing multi-author papers, theses, or complex documents with many cross-references, ExactTeX promises tangible time savings: fewer mystery “??” references, safer renames, and less manual grep-and-fix.
A quick, plain-language note on what “gradual” means here: you don’t get a monolithic new language; you get optional annotations that document the author’s intent. That cuts both ways — it means you can adopt quickly, but it also means coverage matters. ExactTeX reports coverage metrics so teams can decide how much of the document they actually want checked. If you care about full-safety, you can aim for 100% coverage; if you only care about tables and figures, you can annotate those and be confident those items are safe.
Closing Thought
ExactTeX is a reminder that productivity and correctness improvements don’t always come from bigger runtimes or smarter AI — sometimes they come from applying a simple, battle-tested idea (gradual typing) to a different problem space. For academics and technical writers who have wrestled with brittle refactors and inscrutable compiler warnings in LaTeX, ExactTeX is worth a try: it’s low-friction to adopt, brings editor-grade diagnostics, and puts a name-aware safety layer on top of decades of TeX tooling. If enough projects adopt it, the modest improvements in naming and refactor-safety could compound into fewer wasted hours and less brittle collaboration.