philosophy
I barely take photos. And I'm building a photo organizing service
· Ascendy
I don’t enjoy taking photos. I take one or two when I go somewhere, and the rest of the year I take almost none. So photos never piled up on me, and there was nothing to organize. I have no memory of struggling with photo organization.
And yet here I am, building a photo organizing service.
The person who needed it wasn’t me
My wife is the opposite. Food arrives, she photographs it. She buys something, she photographs it as a keepsake. She plans all our trips, and every piece of information and every ticket she needs gets saved as a photo. After our baby was born, she took hundreds of baby photos.
I take photos now too, since the baby. But I really only take baby photos. So my gallery still has nothing to organize. Hers has baby photos and food photos and ticket photos all mixed together.
The person who needed organizing wasn’t me. It was her.
The first ask failed
Last spring we held our child’s first birthday party. When the photos came back, the baby had been crying so hard that every shot was a grimace.
Here’s the part I like: the first person she asked wasn’t me. She asked GPT — just brighten the expression a little. This was before today’s image models, and what came back was a photo where my child’s face had been replaced with a different face.
So the ask came to me. Build an AI that can do simple retouching.
I tried, then gave up. Two reasons, both true. One, I’m not someone who can build an image model himself. Two, even if I could, I judged that this is a game won by whoever has the most data and the biggest computers.
The second ask
So I asked her: is there anything else you need?
She said: build me an AI that organizes photos.
That one felt doable. I’m a developer, and I also wanted to try writing code by directing an AI to do it. That’s how this started.
What I built first was useless
I kept the first version simple. Tags get attached to photos, and you search by those tags. It worked.
But the moment a question got even slightly complex, it fell apart.
“We took the baby on a trip this year — when was that?”
“Out of the beach photos with the baby, show me only the good ones.”
Tag search doesn’t do that. In the end a person still had to look through the photos one by one. I had bolted AI onto it, and the work left for the human was exactly the same. I had gained nothing.
Looking back, that was the most important moment in this project. Giving someone one more search box and doing the organizing for them are completely different things. The first is easy to build, and once you’ve built it, nothing has changed.
So I changed direction
I rebuilt it around what AI is actually good at. Then I fixed, one at a time, whatever my wife said was annoying. That loop kept adding to the project. And once you’re holding on to someone else’s photos, security is something you have to think about too. Building alone, I looked up one day and I was building a photo cloud.
Where it is now
I’m satisfied with where it is. It organizes by time, place, and person, and I find what I’m looking for.
But I have to say this carefully. I’m someone who never had many photos to organize in the first place. My library is an easy problem for this service. My wife seems to be getting along with it too, but that’s as far as it goes. So far this service has essentially been shaped around how my wife and I use it.
People shoot different kinds of photos, and they want to sort them by different things. Someone will want to group by person, someone else by trip, and someone else will want receipts and documents pulled out on their own. That’s not something I can figure out by sitting here imagining it.
So what I need now is for other people to use it and tell me where it gets annoying. My wife’s complaints carried this service this far. Other people are next.
You can see it at ascendy.ai. Photos you put in get organized by time, place, and person, and you ask for what you want in plain words. It runs in the browser. The signup screen asks for an email and a password, and the pricing page has a free plan. There’s no app yet.
If something annoys you while using it, tell me — that’s the information I need most right now.
There’s a separate piece on the same starting point from an engineering angle: Photo clouds solved storage, not finding.
Tags: photo-management, family-photos, why-i-built-this, ai-organizing