The advice is close to unanimous. Pick one job. Do it better than anyone else does. Expand later, if ever.
It is good advice. Most of the software you like follows it. We went the other way on purpose, and I want to lay out the argument rather than just assert it, because I think it is the kind of decision that either explains everything downstream or looks like sprawl.
What a specialised tool cannot see
Take a Tuesday.
There is a task with a deadline on it. There is a note somewhere with the actual thinking behind that task. There is a recurring expense that only makes sense if you know which project the task belongs to. There is a habit you are trying to keep that keeps losing to the deadline. And somewhere in the middle of all of it is another person with their own version of the same Tuesday.
Five excellent apps will hold those five things, each one faithfully. What none of them can hold is the Tuesday.
I want to be fair about why. A task manager that reasoned about your grocery budget would be a worse task manager, and its users would be right to say so. The seams exist because specialising works.
But my problem was never that I had too many tasks. That one is solved, several times over, by people who are very good at it. My problem was that I could not see how any of it fitted together, and the fitting-together was the part I kept failing at.
So the scope followed from that. Boards and cards for the work. Notes for the thinking. Checklists, habits, focus sessions, a weekly planner. Expenses with real multi-currency handling and splits, because money is not a separate life that happens in a separate app. Whiteboards, because some thinking is spatial and refuses to become a list.
The part that gives it away
There is a family suite in here. Chores, allowances, points, houses with a running house cup.
No serious productivity tool ships that. It reads as a different product wearing the same clothes, and I expect reviewers to say so.
It is here because the Tuesday argument does not politely stop at the edge of one person. Household admin is the same species of problem as the sprint board: the chore rota, the allowance, who owes whom after the shop. For a lot of people it is most of the admin they actually do. It usually gets handled with a worse tool, or no tool, or a recurring argument.
Deciding to hold both is what shaped the engineering. Sharing and permissions landed in the second phase instead of the tenth, because a family workspace is useless without them. An expense had to know how to split itself. The data model had to survive a nine-year-old and an accountant looking at adjacent screens without either of them seeing something they should not.
What it costs
A lot, and I would rather say so than let you find out.
Every module is surface area for the next bug. A permission model that has to stay coherent across a board, a note, an expense split and a chore is harder than one that only covers boards. Onboarding has more to explain. Anyone comparing us feature-for-feature against a specialist will find places where we are thinner, and they will be right.
The bet is that the seams cost people more than the depth does. That a plainer notes editor inside something that can see your Tuesday beats an excellent one that cannot.
I cannot settle that by writing about it, and I am not going to pretend otherwise here. What I can do is show what the bet forced us to build: a real database living in your browser so the whole thing works on a plane, a hard split between the data you own locally and the data a server has to arbitrate, and much later an assistant that could read across every module because there was finally one place to read from.
Those choices only make sense once you know what we were trying to hold together in the first place.
About the name
Marginalia. The notes people write in the margin of a book, running alongside the text in the space the author left blank.
I liked that the margin is not the content. It is the record of someone making sense of the content, and it is usually the part worth keeping.