← back to work
~ project no. 01 ~

daisy.

A stylist app I built for myself — to stop guessing in the morning, shop smarter, and feel good in the colors that actually love my face.

2024 → in progress · 0 → 1 · ios · swiftui
~ live preview · tap around ~
This is the real app, ported to web. Not a video, not a Figma frame.
9:41
Good morning
Atara
77°Partly cloudy · H 86° L 68°
Jerusalem, Israel
Tee weather.A light layer for the breeze.
Today's Pick
TOP
Dusty Rose
BOTTOM
Cool Taupe
HAIR
Icy Lavender
Warm rose top against cool taupe — lavender lifts the face on a mild cool day.
Reach for a dusty rose or soft lavender before noon — your skin runs toward cool today.
~ design system ~
~ the daisy design system ~

a hypothetical system, built for every season.

color · soft summer
muted, cool, low contrast
dusty roses ~ primary
50
200
400
600
900
smoky seafoam ~ supporting
50
200
400
600
900
dusty periwinkle ~ accent
50
200
400
600
900
blue-gray ~ neutrals
50
200
400
600
900

for autumn → swap roses for rust, seafoam for olive, periwinkle for warm brown. same ramp structure, different mood. ✿
type · the daisy stack
3 voices, no more
display ~ headlines, herolora bold
soft summer.

body ~ reading, listslora
Aa Bb Cc — these are the colors that love your face. The quiet ones. The ones that make you look rested.

accent ~ tags, captions, asidescaveat
Aa Bb Cc — saved for the little things.
components
a peek
buttons

color chips

garment tiles
linen top
dusty rose · #D6A5B8
seafoam cardi
smoky seafoam · #A6B7B6
motion & soul
the tiny things
loading bloom
spins while ai thinks · 1.4s loop

save sparkle
twinkles when you save an outfit

empty state bow
bobs gently · "your closet is empty"

hand-drawn arrows
replace icon arrows everywhere
the icon language
all drawn by hand
daisy
bow
heart
star
hanger
dress
shoe
mirror
~ the brief ~

why I built this

I was feeling meh about my wardrobe. Nothing was wrong with my clothes — they just weren’t working on me, and I couldn’t say why. I didn’t know color analysis was a thing. I asked ChatGPT to look at a selfie and tell me what colors I could wear, fell down the rabbit hole, and landed on soft summer — muted, cool, low contrast. Once I knew that, the wrong-feeling pieces in my closet finally made sense. Going into 2026 I’d made a resolution: this was the year I was going to take control of my wardrobe and actually feel good in what I wear — not just dressed, but pretty. So Daisy isn’t only about making the morning easier. It’s about feeling nice in the mirror at the end of it. The app I wanted from day one: not a generic palette PDF, but a tool that knows my season, my face, and the clothes I actually reach for, and uses that to take the deciding out of getting dressed.

~ research ~

doing the homework

The research was me, in the mirror, with ChatGPT and Claude open in two tabs. Every color analysis quiz, every YouTube swatch test, every blog post on draping. The conclusion was always the same: the system works, but applying it day-to-day in front of your closet is the hard part. The questions I kept hitting weren’t what’s my season? — they were can I wear this top with this bottom?, does this eyeliner fight my coloring?, is this safe to buy?. Those are the questions Daisy tries to answer. I had Claude work out the safe color pairings inside soft summer for me, then built the guides — makeup, metals, sneakers, dresses — around the clothes I actually wear, not a generic soft-summer reader.

~ ai prompts & evals ~

putting AI to work

Two AI surfaces shipped first. Olive, the in-app chat stylist — ask her anything outfit-shaped, and she answers from your palette, your guides, your closet. And a face-color analyzer: take a selfie, get back the eight or so colors that actually light up your face, not just your season’s stock palette. The analyzer was the harder one. Soft summer alone is ~127 colors across six families, but only a handful really do the work on my features. The prompt evolved through dozens of selfies — different light, with and without makeup — to land on something that picks colors confidently and quietly, and knows when to say this isn’t your color. Next surface: AI-assisted outfit building, where the model pulls from your closet instead of the catalog.

~ ui system ~

palette-swappable at the root

Today the tokens are soft summer; tomorrow they swap to deep winter, light spring, soft autumn — same ramp structure, different mood. Three voices in type: Lora bold for display, Lora regular for reading, Caveat for the little handwritten notes. The illustrations are hand-drawn doodles instead of stock photography on purpose — it keeps the app feeling like a friend, not a catalog. The spec card above is the whole system.

~ motion ~

loading bloom & other soft tells

Motion is doing real work here: a daisy bloom rotates while AI is thinking, a soft sparkle confirms a saved look, an empty-state bow nods at you when your wishlist is bare. Nothing bouncy, nothing snappy — soft summer in motion form.

~ what shipped ~

what’s on my phone today

Sideloaded right now: a home that pulls today’s styling tip, the weather, and a live color picker (eyedropper any photo to identify the color and check it against your palette); a full palette browser (six families, 127+ colors, every safe pairing); face colors tuned to me; style and makeup guides written for my body and my clothes; Olive the chat stylist; an outfit builder; a wishlist; and a journal. What’s next: AI baked deeper into outfit building (pull from your closet, not just the palette), and the universal version — open the app once, AI reads your face, the whole experience tailors to you. The dream is fewer meh mornings, smarter shopping, and a closet that loves you back.