We did not set out to build a product. We set out to stop losing things.
We are a small studio in Christchurch. Between us we were running client work, building software, and running households with school runs, appointments, sport and the ordinary weight of family life. We had all the tools. Calendars, task apps, notes, shared documents, more than one of each. We were still dropping things, and the things we dropped were never random. They were always the ones that sat between two systems.
The follow up that lived in a work tool but depended on a Thursday afternoon that belonged to a pickup. The birthday that was in a personal calendar while the week was being planned from a work one. The thing somebody was waiting on, which was perfectly recorded, in a place nobody looked on the day it mattered.
The problem was never the tools
That was the part that took a while to admit. Our tools were fine. We were the integration layer.
Every evening, one of us was doing the same unpaid job: reading four sources, holding them in one head, and working out what the next day actually looked like. It is invisible work, it is genuinely tiring, and it does not show up anywhere. Nobody thanks you for the crisis that did not happen because you noticed something on a Tuesday.
The reason no tool solved it is that no tool was allowed to see the whole thing. Work software is built for work. Household apps are built for households. Real life does not sort itself that way. A school pickup in the middle of a weekday afternoon is a work constraint. A client call that runs long is a household problem by six o'clock.
The first version was a text file
The first version of this was not clever. It was a plain file describing how a week actually ran. The days that were work days. The hours that were protected. The pickups and their real times. Who was in the household. What we had already decided and were tired of re-explaining. How we wanted to be spoken to, which for one of us was a fairly blunt instruction not to pad and to push back when the reasoning was thin.
We gave that file to Claude and asked ordinary questions. What does this week look like. What is going to go wrong. What have I not replied to.
The answers changed. Not because the model was different, but because it finally knew enough to answer for us rather than for nobody. It stopped suggesting a Wednesday afternoon. It started noticing that two things we had both agreed to were the same two hours.
That was the whole discovery, and it is smaller than it sounds. Context beats cleverness. A capable assistant with no idea who you are is a stranger. The same assistant with one honest page about your life is something else entirely.
What we added next
From there it grew in the direction real use pulled it.
A briefing, because reading the week in one place was the thing we did every day. Saved instructions for the jobs we kept redoing the same way, so we stopped re-explaining how we wanted the weekly read put together. A view of the week we could look at rather than ask for. Somewhere for the loose ends to live so they were not stored in one person's memory.
None of it was ambitious. Every piece got added because something had gone wrong once and we did not want it to go wrong again.
Why we turned it into something you could set up
For a long time this was just ours. What changed our minds was watching other people try to get there.
They were not short of capability. Claude was available to them, and mostly they had already tried it. They were short of setup. They did not know what to tell it, so they told it nothing, and got generic answers, and reasonably concluded it was overrated. The gap was never the model. It was the hour of honest work that turns a general tool into your assistant, and almost nobody does that hour on their own, because it is not obvious that it is the whole game.
So we built the thing that walks you through it. A questionnaire that asks the questions we had to work out by trial and error. A setup that writes your file from your answers, on your own Claude account, with your details staying yours. A first saved skill that is already there and working, so you can see how they work well enough to build the next one yourself.
What we deliberately did not build
We did not build a subscription you rent from us. Everything you end up with is a plain file on your own account, and what that includes is set out in full rather than left vague. If we disappeared, your setup would keep working, which felt like the only honest way to sell somebody a setup.
We also did not build another place for your information to live. There are enough of those, and adding one would have recreated the exact problem we started with.
Where it is now
It is not open to everybody yet. We are letting people in a few at a time, because the part that matters is whether the setup actually works for somebody at the end, and that is not something you can tell from a signup number.
If any of this sounds like your kitchen table, it will probably sound like ours too.
Deb and Zaskia, Orchid Horizon