Building FamilyOrganization.app in public

When we tagged every comment, we expected a list of features. Instead, 12 comments mentioned Android and 6 asked about TVs other than Roku or said they don't have one. Four asked about Alexa, three about Fire TV, two about Apple TV and one about Google TV. The feature requests were smaller and mostly already built: calendar sync from the phone, countdowns for time blindness, a way for kids without phones to tick things off.

So the gap wasn't features, and counting every comment, rather than answering the loudest one, is what showed us that. In the two weeks after the posts we wrote an Android version of the app, researched three more TV platforms, and made the backend pair any of them. None of those are released yet, and we'd rather say "coming soon" than pretend.

Nobody wrote "your signup form rejected my phone number". It did, though. The form treated any 10-digit number without a country code as American, so a Dutch mobile number, typed the local way, became a broken "+1" number and failed the whole signup with a generic "try again later" message. The person saw an error and, most likely, gave up.

We fixed it on 15 September with a country picker and proper per-country parsing. The commit message says it plainly: "Real international signups were being silently lost." When a post brings strangers, the comments tell you what they hoped for, and the pipeline tells you where you're losing them. Check the pipeline first, because nobody complains about a door they couldn't open.

Once families in the UK and Australia asked to test, time itself became the product problem. Our routines, quiet hours and chimes are wall-clock times: 7:00 is 7:00 for the family, whatever zone the TV is set to. A TV set to a different zone ran the family's morning at the wrong hour. And a lot of the world doesn't read a 12-hour clock or start the week on Sunday.

On 21 September we added a 12- or 24-hour clock for the whole household, a Sunday or Monday week, and a question the TV asks once when its zone differs from the family's: which time should this TV use? We tested it on our own TV by pretending we lived in London. It's a small setting, and it's the difference between an app for one country and one ready for families in twelve.

The most-reacted comment in the whole export suggested dropping the AI. Others asked, fairly, how safe their family's information would be. We kept working on the AI connector, because telling Claude "add these three chores for tonight" and seeing them on the TV is one of the things that makes this useful in our house. It's still in development, it will always be optional, and the app works fully without it: AI isn't involved unless you want it to be.

In a pre-launch privacy review that same week, on 21 September, the connector stopped being able to read the family's home address, only a household admin can connect one, and every approval is now recorded. Each request already ran as the signed-in parent, so it could never see another family. Trust isn't a feature you announce once. It's the list of things the app won't do.