Furrell builds the unglamorous things a multi‑site organisation runs on. The current one is Onsite — every room, every session, every building, and the one question a spreadsheet can never answer.
When organisations merge, their buildings merge with them — and several hundred people suddenly need to know which building they are in, which room they have, and whether it is still theirs next Tuesday. Onsite is the screen that answers it — and the place shifts get published, claimed and approved, with a reliability record that surfaces while somebody is deciding rather than in a report nobody runs. Every organisation sets up its own estate, and sees nobody else's.
Northgate · Meridian Group
Offering 13:00–17:00 — “Happy to take this one but I have to be away by five.”
Pulled out of 1 confirmed shift in the last 14 — 7%, below your 20% flag.
Working reconstruction · invented data — try both tabs, and the buildings
Separation between organisations is a row-level security policy in Postgres, not a filter somebody has to remember. A screen with a bug returns nothing, rather than someone else's rota.
One person in two rooms at once is refused by the database itself. Two people in one room is only warned about — because a trainee sitting in alongside someone senior is not a mistake.
Rooms retire, people deactivate, bookings cancel. Last March's rota still points at every one of them, so it still reads correctly a year later.
A different kind of thing entirely, and the one that proves the rest: designed, built, submitted and shipped alone, through App Review, without a team to hand it to at any point.
Book of Intentions
New moon · 8 September
“A new moon tonight — a dark sky for asking. Pen what you want next.”
An intention book that keeps the year's turning points — moons, festivals, the quiet weeks between — and asks at the right moments rather than every morning. It writes in two hands: Midnight, the book by candlelight, and Parchment, the same book by daylight.
Every word stays on the reader's own device. There is no account, no server, and nothing to leak — which is a design decision before it is a privacy claim, and it is the reason the app needed no backend at all.
Onsite did not begin as an idea. It began as six internal tools built for a group that had merged faster than its systems had — which is how the problem got understood well enough to build a product for it.
A whole team modelled against what the clinics actually need. Its one finding turned a recruitment round into a training plan.
Reads the plan, checks it against what's rostered, and turns the gaps into actions before the phones open.
Hundreds of thousands of triage records against an ideal line a board can read in a glance.
Projects, unclaimed ideas and quarterly priorities — the three lists that never agree — in one place.
The absence file every team already has, except built properly. In use by teams I have never met.
The score everyone quotes and nobody explains. Two numbers in, the number and its meaning out.
Open any one for the full story and a working demonstration — all data shown is invented.
Most of the tools here started as somebody's frustration.