Word makes tracked review familiar. Overleaf makes equation-heavy collaboration practical. GitHub preserves version history and supports reproducible checks. Each solves a real part of manuscript development.

The friction appears when the same draft repeatedly moves among them.

My current experiment reverses the usual order. The everyday manuscript remains readable and directly editable in a web interface. LaTeX and PDF become controlled outputs produced at submission checkpoints rather than the only place where the argument can be revised.

What the workflow must preserve

A useful implementation cannot trade scientific control for convenience. It must preserve equations, references, figures, tables, and document structure. Export should be deterministic, revisions should remain visible, and the generated LaTeX should compile without silently changing the scientific content.

Confidential manuscripts and unpublished data must also remain in an appropriate local or private repository. No automated revision is accepted until an author reviews the resulting argument, evidence, and wording.

Why test this approach

The purpose is not to prove that a web editor is universally better. Different coauthors have different habits, and tracked changes in Word or direct editing in Overleaf may still be the right choice.

The narrower question is whether a web-first draft can reduce the repeated editing, compiling, pushing, and pulling that interrupts scientific reasoning. If it can, the saved effort should be measured in fewer synchronization errors and faster substantive review, not merely in the number of automated steps.

Current status

This is a working prototype, not a released manuscript platform. The next tests concern citation round-tripping, equation fidelity, reviewer-style diffs, reproducible LaTeX export, and coauthor usability.

The short film introducing the idea is available in the featured-film section on the homepage and in the original X post.