Building in public

Marginalia

Notes from the margins. Short, unpolished, written as we go ... the half-formed reasoning behind what we're building.

Tuesday
13:35

when you sit down to write, you don’t always know what the thing is yet. an outline? a document? a few bullets that turn into three paragraphs that turn into a chapter? most apps make you answer that before you’ve written a word. they’re either an outliner with no real prose, or a prose editor with no real outliner. so you pick the mode, choose the app. and that small decision, made every single time, is enough to stall the writing before it starts.

writing is hard enough on its own … the tool shouldn’t make it harder. our goal with vellum is to remove the friction. build one app that combines the best of a powerful outliner and a gorgeous prose editor, plus typed tags and bidirectional links so any piece of writing can connect to any other, wherever it lives. that way you can just write, however it wants to come out. perhaps start in an outline and convert it to a document, or start as a document and convert it back later.

with vellum we wanted documents to be a lovely writing canvas, not hidden bullets or a view bolted on the side, but a surface that takes prose as seriously as structure. a document is a node, like everything else in vellum, so it can be nested, and it can hold documents of its own. a book and its chapters. a screenplay and its scenes. a storyboard. the whole structure living as real documents inside documents, every one writable, every one a thing in its own right.

because to us a document is just a node. so it nests like an outline and connects like a graph, both at once. one home in the tree, and any number of links out to anywhere else.

we’re not trying to build a filing cabinet. we’re trying to build the kind of place you actually want to sit down and write in.

Thursday
22:35

a question came up in Discord today about syncing and end-to-end encryption. the answer is really just the whole thesis again from a different direction.

most apps solve sync by making the cloud the real copy. your notes live on their server and your devices are just windows onto it. convenient, and also exactly the thing we’re trying to get away from, your stuff sitting readable on someone else’s machine.

the model we want is the one that Bear and a few others use: every device keeps its own complete copy that works fully offline, and edits flow between your devices in the background and merge. no central version. sync with yourself, not into a cloud.

that’s the hard way. two devices safely touching the same data, conflict handling. it’s a fair bit of work for an app like this … so we’re building it as its own deliberate piece rather than bolting it on.

and once data syncs, end-to-end encryption is the obvious end state. everything starts local, so your data is already yours, and e2ee keeps it that way wherever it goes. no readable copies anywhere but your own devices.

Tuesday
09:40

the core thesis behind vellum:

  • notes should be owned and solely accessible to the note-taker, not a SaaS platform
  • notes shouldn’t be trapped inside an app. If the app dies or the company folds, then what? Apps are fungible, notes are not.
  • every node (bullet, paragraph, code-block, quote, equation, image, etc.) should be easy to capture, discover, connect, zoom into, and build on
  • last but certainly not least, a notes app should be genuinely lovely to use … the digital equivalent of a good fountain pen on fine paper (minus the drawbacks of handwriting)
07:10

thanks for being here.

this is marginalia … notes from building vellum, thinking out loud as we go. feature ideas, design decisions, the occasional dead end. nothing polished, nothing on a schedule.

if you’d like to follow along for the ride … we maintain a small, private Discord. a handful of early folks and us, talking through what’s being built, what’s broken, and what’s next. it’s invite-only and deliberately cozy … if that sounds like your kind of room, leave your email and we’ll send you an invite.