An office suite whose documents are HTML
Open a document in a browser and it renders. Open it in a text editor and it reads. Put it in git and the diff is readable. No converter, no runtime, no format anybody has to reverse engineer in ten years.
Why it is built this way
A browser engine is already a layout engine, and the most heavily optimised piece of software most people run. It handles text shaping, line breaking, tables, floats and fonts. Word processors normally reimplement all of that, which is why there are about three of them in the world.
So the document is HTML and CSS in a zip, and the pages are computed instead of stored. A change of margin does not rewrite every paragraph, and a PDF agrees with the screen by construction.
What is actually in a file
A zip, whose first entry is an uncompressed mimetype so a tool can tell what it is holding without unpacking it. Inside: the document as HTML, the stylesheet that lays it out, and the pictures and fonts it needs. Nothing generated, nothing cached, nothing that only our software can read.
The extensions are .aupg, .aush, .audk and .aubs: au, and two consonants.
What an account is for
The apps work offline and keep working. An account covers the whole suite and buys the parts that run on our side: storage, sharing, a relay. Nothing that happens on your own machine is behind a tier, because a word processor that stops opening your own files when a card expires is not a product anybody should build.