The problem
I run my business and my life out of Obsidian. The planners I tried were either spreadsheets with a calendar glued on, or apps that wanted me to live inside them. I wanted a daily page that felt like a paper notebook: open it, write, close it. Tasks, a weekly spread, a monthly view, and handwriting on every surface. On a Mac, and on an iPad with a pencil.
Constraints I chose
Markdown stays the source of truth, so nothing I write is trapped. Every view has to answer four questions without a manual: where am I, what's here, what's next, how do I get back. And the chrome stays quiet. A page you write on should not look like a dashboard.
How I make decisions
Screenshots. I use the product all day, and when something feels wrong I take a screenshot on the device where it felt wrong, write one line about what bothered me, and that becomes the brief. The before and after get kept next to each other. When a change takes three screens of pills to explain, the change is wrong.
One recent example. The daily page had started carrying full editing chrome for every embedded page: tools, page navigation, a side rail, bottom nav, a zoom bar. Three screens of controls on a page meant for writing. The decision was to show the ink and a single Edit chip, and bring the chrome back only when you ask for it. The page got quiet again and the writing came back.
A design system you can't drift from
Oak has a written design law: colours, type, spacing, radii, the rules for quiet chrome. I turned it into a linter that runs in the build. It counts the debt in every stylesheet and fails the build if the count goes up. Design drift stops being a review comment and becomes a failed build. The same idea runs the design system on this site and on the SEMA properties.
Touch, designed for a pencil
Altitude is the ink layer: an infinite canvas where I write and draw with an Apple Pencil. That forced the problems a mouse never shows you. The side of your hand lands before the pencil does, so palm rejection has to be right or the page is useless. Pinch to zoom fights two-finger scroll. A toolbar that is fine at a desk is in the way on a lap. Each of those got a screenshot, a decision, and a live check on the real iPad before it shipped.
Altitude also learns your handwriting on the device. You write a few lines with the pencil, nothing leaves the iPad, and recognition gets better for your hand specifically. The training screen had to feel like a notebook page, not a settings form.
Shipping it
Every change goes through a strict build (types, design lint, bundle), a test suite, and a live check that drives the running app and proves the installed build is the one I think it is. Releases are one command. That discipline came from a day I shipped a build from a dirty tree and quietly reverted a week of someone else's work. I wrote the gate the same afternoon.
What I'd do differently
Start on the iPad. I designed on the Mac first and ported to touch, and most of the rework came from that order. The small screen with the pencil is the harder constraint, so it should set the rules.
