In our Oakland neighbourhood, Esther started helping neighbours coordinate group rates on the services everyone was buying anyway, cleaning, pet grooming, car detailing. She ran a vendor spreadsheet covering around sixty homes: calling providers, working out who was good and what the right price was, then batching schedules so a groomer could handle several households' dogs in a single visit.
The spreadsheet broke when word travelled. People as far out as San Jose started asking to join, and that was the moment it stopped being a favour and started looking like a business. My first contribution was a website so people could book and pay themselves, with payment routed through to the providers.
Then the usage data disagreed with the customers. The services people told us they wanted were often not the services they actually booked, and that gap shaped the second version. A strategic shift produced the third and current one, which is concierge-led: a client app, a concierge app, a service-provider app, a provider admin panel, a main admin panel and the backend tying them together, with AI helping a single concierge manage many homes at once. There is also a building-benefit model now, so residential buildings can offer Geoffrey to their residents.
Geoffrey was Esther's idea. For a long time Esther was also the product. I built the software because the alternative was watching her keep doing all of it by hand.
Memory that respects 'forget that.' An early build dutifully remembered things users had explicitly asked it to forget. The fix, tombstone everything, and assume every cache layer lies to you at least once a quarter, is now a rule I apply to every system I touch.
Live and doing well, growing neighbourhood by neighbourhood and building by building.