Back to Main

Blinkko Labs

This case study is password-protected. Enter the password to view it.

Opportunity

Blinkko is a shoulder-worn social AI companion — hardware plus App — for urban young people. The product should turn going out into exploring the city, making a connection, and showing who you are.

I joined as founding product designer in June 2026, owning software design with the software team: user research, product strategy, information architecture, the UI system, interaction, and shipping the App alongside the hardware for the September launch.

The problem Functions were scattered. Opening the App felt like configuring a device, not meeting a companion you wear in public.
The job Rebuild the App around real use: home as the hub, five spines, software and hardware in one loop.

Early traction

1,900+ Waitlist signups
658 San Francisco launch RSVPs
7.6% Meta ad CTR
$0.30 Cost per click · as of 2026.09.05
Blinkko product render
Blinkko at night Blinkko at the beach

Insight

If other people can read the glow and the touch, the opening stays low-risk — so you keep wearing it, and it keeps sensing. Wear, sense, understand, respond: that loop is what the whole product is built on.

Research with Gen Z pushed the interface away from settings-speak, toward the way they actually talk. The first session should introduce a companion, not configure a device.

System

From information architecture to a shared component library. After I took over, I folded scattered functions into five spines and a shared component set, and led both. The shipped App is four tabs — Explore, Home, Friends, Profile — with Home as the hub. Hardware is optional: light, bump, and three-tap check-in appear when a session is live.

Workflow

First drafts come from a skill the team already had, which builds a PRD out of design-system components. My time goes where judgment matters: the flow first, then the pixels.

  1. PRD lands Written by the founder.
  2. Skill drafts the UI A first pass, built from design-system components.
  3. I take over I rework the flow first, then the pixels.
  4. Launch Gaps filled screen by screen, on the launch schedule.
Tools used Claude Code Claude Cowork Figma MCP Lovart Codex

Principles

01 · Type is the interface Large, bold, spoken lowercase (“sup, kai”). That came from looking at how Gen Z actually talks — the type is the UI, not decoration on top of it.
02 · Dark base, light as signal Dark UI lets the character, lights, and Glow take focus, echoing the companion on the shoulder.
03 · State stays in sync Light, volume, bump, and Send Glow have a place in the UI because they exist on the hardware. Tap Glow, and the other person’s tail lights up.

Information architecture

Home as the hub for five spines. Onboarding leads into the Home hub; from there Homepage, Friend loop, City Explore, and Pet growth are the four you return to every day. Settings sits beside them as a utility, not a spine. Mapped in full, the five spines come to 96 touchpoints; the diagram shows the top level.

Five loops

Onboarding, Homepage, Friend loop, City Explore, Pet growth — in that order. Screens below are the landed UI, autoplaying; tap to step.

Onboarding

Welcome: “a light born from what you feel…” then let’s go. Sign-in is Apple / Google first; avatar and username finish the free-app path. Once you are signed in, the vibe quiz lets Blinkko get to know you. The permission stack is no longer required up front.

Welcome → sign in

Deep dive · Onboarding

Onboarding had to work for three kinds of people who arrive with very different levels of commitment: someone who has just unboxed the hardware, someone who came in from a quiz on social media, and someone who found the App in the store.

AHardware owner

Just unboxed the device. High commitment, so they can be asked more.

  1. Unbox the hardware
  2. Download the App
  3. Welcome + sign up
  4. Permissions
  5. Bluetooth pairing
  6. Meet Blinkko (ceremony)
  7. Full quiz, 5 questions
  8. Name + customize
  9. Home

BSocial media funnel

Arrives from a quiz ad. No device yet, and may never get one.

  1. “What Blinkko are you?” quiz ad
  2. Mini quiz, 3 questions, no login
  3. Result preview: “sign up to keep”
  4. Download the App + sign up
  5. Full Blinkko reveal
  6. Name + customize
  7. Pair a device, or skip
  8. Home

CApp store discovery

Found the App on their own. Needs a reason to stay before a form.

  1. Hook slides
  2. Mini quiz, 3 questions
  3. Blinkko preview
  4. Sign up: email, Apple, Google
  5. Explore without hardware
  6. Pair later, from settings
  7. Home

First version

The first version was designed around the hardware owner: 14 screens in three phases, each with a different job. I had the whole sequence wireframed and tested its pacing before any visual design.

Phase AFunctional setup

As fast as possible. Don’t annoy people.

Wireframe: Welcome Wireframe: Sign up Wireframe: Permissions Wireframe: Bluetooth pairing

Phase BThe encounter

The product’s first impression. It has to make people go “wow”.

Wireframe: Transition to darkness Wireframe: A faint light appears Wireframe: First touch Wireframe: Blinkko takes shape Wireframe: Blinkko fully awakens

Phase CBonding

Building the relationship. It should feel warm.

Wireframe: Introduce yourself Wireframe: Get to know each other Wireframe: Name your Blinkko Wireframe: Customize appearance Wireframe: First light together

What the review surfaced

Hook before sign-up Show the most interesting thing first. Don’t open with a form.
Not everyone has a device Paths B and C start without hardware. A flow gated on pairing stops those people at the door.
Two “first meetings” Pairing and the quiz each reveal a Blinkko. Doing both made the encounter feel repeated.
Delight over thoroughness For a consumer product the first minutes should feel light, not like a questionnaire.

Testing three ways in

The review left an open question: what should a new person be asked to do first? I took three answers to that question to 30+ target users as three onboarding schemes: one led by the purchase, one led by a free look around before owning anything, and one led by a personality quiz.

30+ Target users in the sessions
3 Schemes: purchase-led, free-experience-led, quiz-led
+30% Participation conversion of the quiz-led scheme over the other two, in testing

The quiz-led scheme

In testing, the quiz was the first thing a new person did. In the shipped build it sits one step later: sign-in is a single tap with Apple or Google, not a form, and the quiz follows straight after, so the result belongs to an account from the start. It is the moment Blinkko gets to know you, and the first place the App becomes personal: two details about you, six questions written the way the audience talks, and a result with a card made to be shared.

Quiz: blinkko wants to know about you
About you
Quiz question: it’s friday 6pm, what’s the move
Question 1 of 6
Quiz question: your camera roll is mostly
Question 3 of 6
Quiz question: which one actually hits
Question 6 of 6
Shareable result card: night owl
A card to share

The result: one of five vibe types

Vibe type: homebase
homebase
Vibe type: night owl
night owl
Vibe type: side quest
side quest
Vibe type: hidden gem
hidden gem
Vibe type: spotlight
spotlight

Hardware stayed optional throughout: pairing moved out of onboarding, so the map and the social features work before anyone owns a device.

Homepage

Hallway greetings from the App — “sup, kai”, “yo, kai”. Light and volume open as overlays on that same character, not separate settings pages.

Home → light overlay → volume overlay

Friend loop

Chats, Send Glow, then hardware Bump: two devices meet, “a new encounter,” then the contact sheet.

Chats → Send Glow → Bump

City Explore

The map tab: discovery, events in town, Glow create, posts, and phone check-in. Everything sits on one map, not a second app.

Map → events → create Glow

Pet growth

The companion grows in the App in two places. The Lumi shop for outfits: preview, spend, confirm, owned. And the monthly recap, where the tags from your month drop onto the city.

Shop → purchase → owned

Report intro → tags drop in

UI component library

The shared library in the file is four groups — foundations & Glow, controls & navigation, cards & feedback, social & companion. Shown here in the order the system is built: Type → Character → Core UI → Cards → Home → Place → Friends. Used across onboarding, Home, Friends, Explore, and Glow. Open a board to see it at full size.

Hardware–software loop

The launch tested that loop end to end: what the shoulder shows, what the device senses, what the App understands, and how both respond.

HARDWARE On the shoulder SOFTWARE In the App 01 Wear A magnetic pad holds it on the shoulder 02 Sense Touch, proximity and motion 03 Understand Place, people and events 04 Respond The App sets light, haptics and the tail Feedback up The tail lights up Signals down What the device senses Others read the light and the touch → a low-risk opening → keep wearing
01 · Wear A visible companion on the shoulder.
02 · Sense Touch, proximity and motion.
03 · Understand Place, people and events.
04 · Respond Light, haptics and the tail.

What we learned

On site it was clear: plenty of people wanted to wear it and change the light. But ask what they could play right now, and almost everything came from users themselves — there wasn’t enough to play early on.

Since launch I’ve pushed a single-player, gamified exploration to MVP, setting requirements, schedule and sign-off with game design, marketing and engineering.

How the team worked

Three teams, one product. Each side owned a clear contract, so the shoulder and the App always told the same story.

Hardware Form, sensing, battery and the light module — the input and output the device can deliver reliably.
Software product design (me) Interaction contracts, state alignment and UI — translating those limits into what the shoulder and the App each do.
Engineering Connection, latency and device-state sync — so bump, battery and light state are the same thing on both ends.

Launch

Blinkko launched in San Francisco in September 2026. Once the software was ready, I flew back to the US from China with the team and ran demos on site, showing people how the App and the companion on the shoulder work together.

Blinkko team
Visit Site