Kettle
Group dinners, planned in four taps instead of forty messages.
Umbrella builds consumer and AI products end to end — the problem, the design, the code, the launch — then hands each one to a founder who runs it. Two are in build. Nothing has launched yet, and this page will keep saying so until something does.
First launch targeted Q2 2027
A studio with nothing launched has only its judgment to show, so we publish the whole pipeline. Killed ideas stay on the board with the reason and the week we stopped.
Group dinners, planned in four taps instead of forty messages.
A reading companion that argues with the book instead of summarising it.
Three incumbents already, and none of them are bad.
We couldn't find anyone who would pay for it twice.
Every tester wanted a different app underneath the same name.
Real problem, wrong team to solve it. We'd have been guessing.
+ 10 more, listed on request
Kettle is the first one out of the studio. When it ships, this column stops being empty and the number above changes.
Targeted Q2 2027Every venture runs the same clock. If a project slips past a mark twice, we close it and the idea goes to the killed column.
We start from a problem we've watched go unsolved, not a market map. One page, one owner, one reason it hasn't been built yet.
Two engineers and a designer ship something a stranger can use. No deck exists yet, and nobody has written the word "platform".
A real listing, a real price, real reviews. We learn from installs and refunds, not from twelve more user interviews.
A founding CEO takes the company and raises on their own terms. Umbrella keeps 25%, one board seat, and stays out of the product.
No venture has reached week 26 yet. Kettle gets there in March 2027 — that date is the first real test of whether this clock holds.
Three designers across the whole studio. A product has a name, a mark, and an icon before it has a database.
Auth, payments, push, and offline sync already work on iOS, Android, and web. A venture starts where a normal build reaches week 12.
Routing, evals, and an inference budget we've already tuned. You choose the behaviour; the plumbing is someone else's problem.
The audience we built at our last companies — a 60K-reader list — and the creators we've worked with since 2021.
Enough to reach the store without raising. No deck, no data room, and no board until you want one.
Payroll, incorporation, contracts, App Store Connect, support tooling. You never file a form on your own.
We start two companies a year and the next slot opens in January. Send a paragraph — not a deck — about the thing you haven't been able to stop thinking about.