The best gifts come from knowing someone. I built Valet to make more of them.

Premise

A good gift starts with paying attention

Most gifting tools start with a list: a gift card, a wishlist, or something someone already asked for. Valet starts with the details that make someone specific, then turns them into gifts, outings, and reasons to spend time together.

◆ Engine

Interests have gears

Someone can love music, make music, or buy synths to learn how to play. Those are three different relationships with the same subject. Valet had to tell the difference.

A point of view, a recommendation, and what happens next

←01 / 03→

The profile makes what it has learned about someone readable.

◆ Voice

I taught Valet how to talk like it knows you

I wanted Valet to sound like a friend who remembers the small stuff. Not “the perfect gift for the foodie in your life,” but “she knows her soba from her udon.” I put that voice into the product so the reason behind a recommendation feels like part of the gift.

◆ Plans

One sentence can become an afternoon

Valet uses the same memory that finds a gift to build a day around someone. “A quiet dinner for Sarah’s birthday, $50 to $75” becomes a plan with places to go, things to do, and a reason for each stop.

From a few choices to a plan made to share

←01 / 03→

Start with the person, the moment, and a few editable details.

◆ The app

A quiet tool that never says its own name

Valet is built around four quiet places: what to do next, a daily question about someone you love, your plans, and an activity feed. You tell it what you know by talking or tapping through a tray. A profile keeps the useful details and the gaps worth filling.

The small details that keep a person in view

←01 / 03→

Roll call gives you one useful question when you have a second.

How it's built

The product changed as I built it

I started Valet with a Figma design to get the idea moving. Once I began connecting real services, the product got bigger: shopping APIs made recommendations concrete, place data opened up Plans, and model and transcription APIs made memory usable. Paper became my sketchpad for working through those changes quickly. I built each idea into a release, tested it, and let the product tell me what needed to happen next.

◆ Agent

I gave Valet a faster front door

An assistant should help in the moment. I reused Valet’s backend to stand up a text frontend, so you can text what you need, go back and forth with it, and get a recommendation or plan without opening the app.

Where it stands

The work is shipped. The proof is next.

The engine runs, the first picks are live, and Valet is in TestFlight. The text version is in internal testing. I built the system far enough to use it, test it, and see where it still needs work.

Receipts

What it actually took

I tested Valet hard. The project logged 407 commits over 69 days, with no reverts. Here is what that work looked like.

Numbers as of July 2026.

The whole company is the distance between “a gift that keeps on giving” and “she's past the grocery-store roast.” One of those could be written about anyone. The other one gets your sister. I'm building the system that reaches the second line every time.