It wasn't just our house

A few weeks ago I wrote about how groceries kept causing friction at home, and how I built an agent to take the thinking out of it: We kept fighting about groceries.

I expected a few polite likes. What I got was more interesting. A lot of people recognised it: in their house, too, the planning quietly lands on one person. And quite a few asked the same question. Can I use this too?

The honest answer was no. My version runs in Cursor, reads files on my laptop and shows its menu on a Home Assistant tablet in our kitchen. It works well for us. It's not something you can hand to someone else.

So I started on a version that you can.

From my setup to an app

The idea hasn't changed. Every week you get a menu proposal, the app works out which basics are probably running low, and it prepares your Picnic cart. You check the cart and place the order yourself. The app never orders on its own and never picks a delivery slot.

What has changed is everything around it. No Cursor, no Home Assistant, no files on my laptop. Just an app on your phone.

The hard part turned out not to be the technology. My version knows our household because I taught it, one correction at a time. A new household starts with nothing. The app has to learn in a few minutes what my agent picked up week after week.

Getting to know a household in five minutes

That's why most of my time so far has gone into the first few minutes: the onboarding.

Instead of a long form, it asks a handful of questions that each tell it several things at once. Who eats with you. What a normal week looks like: busy evenings, cooking every night, or keeping it simple during the week. How you like to eat in the evening. Three flavours you enjoy. Whether price or organic matters more to you.

Then you link your Picnic account. Your order history turns out to be the strongest signal there is, because it shows what you actually buy, not what you think you buy. At the end, the app tells you how it sees your household, and you can correct it before the first week starts.

What a week looks like

The week works the same way as at home. You get a few dinners to choose from, keep the ones you feel like and swap the rest.

Then the app adds the basics: milk, bread, eggs, the things a meal kit never covers. It bases that on your order rhythm and on what you tell it. Mark the coffee as still enough, and it skips coffee, whatever the rhythm says.

Finally, one tap fills your Picnic cart. You open Picnic, check it, choose a delivery slot and place the order. After dinner you can rate what you ate, and those ratings shape the weeks after.

The screenshots use demo data, and the food photos are placeholders for now.

Where it stands

Everything above works from start to finish on my laptop. The next step is moving our own household over, so we're the first to depend on it. If it survives a few weeks of our Sunday evenings, it's ready for other households.

Two things I want to be upfront about. Picnic has no public API, so the app uses the same unofficial connection that other open-source Picnic tools use. It isn't made by or affiliated with Picnic, and something may break when Picnic changes things on their side. And your Picnic password is never stored. The app only keeps an encrypted login session.

Looking for a small test group

This is where I'd like some help. I'm looking for a small group of households who want to try the first version and tell me honestly what works and what doesn't.

In practice, that means:

• you order your groceries at Picnic in the Netherlands (the first version is in Dutch)
• you use the app for a few weeks instead of putting your order together yourself
• you tell me what's annoying, confusing or missing, in a few words or a long rant

I don't have a date yet. I'd rather start when it's ready than rush it. If you sign up, I'll email you when there's something to try, and that's it.