First-time user
Learn the feature quickly
- What do I say, and when is Buddy listening?
- What happens after the photo?
- Where will I find my data?
Meets it throughThe two-page intro and Try Now, a guided first log
01Overview
Nutrition is an AI feature on B Glasses. A user looks at their plate and asks Buddy "What's on my plate?" or says "Hey Buddy, log my meal!". Buddy identifies the dish through the glasses camera, breaks down its nutrition, and with the user's confirmation logs it to a Nutrition card in the companion app, with daily totals and history by date.
Phase 1 at a glance
2User stories
<5 sPerceived latency
11Values per meal
1 yrMeal history
02Where it started
The brief came as a Phase 1 PRD with three pieces to build: a look-and-tell answer to "What's on my plate?", a voice command to log a meal, and a Nutrition card in the app with daily totals, history by date, and delete.
The PRD named two goals, and every design decision had to serve both:
- Visualise health insights. Close the gap between eating and knowing, using the camera and AI to give immediate, actionable data.
- Increase glasses usage. Give people a daily, habit-forming reason to use the glasses beyond music and calls.
03The problem
Eating happens fast; knowing what you ate takes effort. On a phone, you search a food database, pick a portion and save. By then the meal is over, so most meals pass with no insight at all.
User problem
B Glasses users had no hands-free way to understand a meal before the first bite, or to keep a daily record without stopping to type.
Business problem
Glasses usage centred on music and calls. Without a daily, personal use case, there was little reason to talk to Buddy through the day.
How might we…
- Turn a glance at a plate into nutrition data in under 5 seconds?
- Keep users in control, so nothing is logged without their confirmation?
- Make meal logging a daily habit that brings users back to B Glasses?
04My role
I led Nutrition as design lead for B, working with the B design team of one UX and two UI designers. I owned the experience end to end, from reading the PRD to the screens and rules engineering built from.
- Requirement review. Read the Phase 1 PRD against real use and mapped every requirement to a visible design response.
- Discovery and scenario mapping. Scanned existing nutrition and smart-glasses experiences and mapped first-time and regular-user scenarios.
- End-to-end user flow. One flow covering onboarding, the landing page, dates and delete, with every decision point.
- UI design. The first-time experience, every state of the Nutrition landing page, meal detail, history and delete.
05Research and discovery
Discovery combined three things: a requirements review of the Phase 1 PRD, a scan of existing nutrition and smart-glasses experiences, and scenario mapping for first-time and regular users. The scan showed every current option breaks down at the same moment, the meal itself.
| Approach | How it works | Gap |
|---|---|---|
| Manual phone trackers | Search a food database, choose portion, save | High effort per meal; easy to abandon |
| Phone-camera AI trackers | Snap a photo with the phone for an estimate | Phone must come out mid-meal |
| Glasses look-and-tell | Ask the assistant what it sees | Answers, but no structured nutrients or history |
| B Glasses Nutrition | Voice + glasses camera + app card | Hands-free capture, confirmed log, daily history |
Insights that became design principles
01
Hands-free first
The meal is the moment. Capture starts with a voice command, not a phone.
02
Confirm, don't assume
Ask when an item is ambiguous, and never log without explicit consent.
03
Glanceable, not clinical
Lead with calories, then macros, then micros on demand.
04
Earn trust
Use exact brand data when known, and say clearly that values are AI-generated.
06Competitive landscape: direct and indirect
Food recognition from a photo is fast becoming a commodity. The real competition is over when capture happens, who stays in control, and what brings people back. That reframing shaped Phase 1 more than any single feature did.
How I mapped the market
I didn't want a feature checklist. I mapped competitors in two rings and judged each one through the same five questions a user lives through at the table:
01
Capture
Does the person have to stop eating to log?
02
Control
Who confirms what gets saved, the AI or the user?
03
Uncertainty
How does the product behave when it isn't sure?
04
Context
Does it understand a multi-dish Indian plate?
05
Return
What brings the person back tomorrow?
Direct competitors solve the same job on the same form factor: AI glasses that log food. Indirect competitors solve the same job another way: phone-camera AI trackers and manual logging apps. Our users already know these, so they set the bar for expectations.
Direct: AI glasses
Ray-Ban Meta and Oakley Meta. The closest benchmark. In April 2026 Meta announced food logging by voice prompt or quick photo, with Meta AI extracting nutrition details into a food log in the Meta AI app (9to5Google). It launched for US users aged 18+, and users still prompt Meta AI and set nutrition goals themselves; automatic logging is promised for later (Road to VR).
What it told me
Hands-free capture alone is not a moat. The global leader is moving the same way, so B has to win on the experience around capture: confirmation, clarification and local context.
Indirect: phone-camera AI trackers
| Product | How capture works | What stood out for design |
|---|---|---|
| HealthifyMe Snap | Phone photo; trained on 150,000 Indian food items (TechCrunch) | Built for Indian plates. With several items in one photo, users select and track each in an adjustable box (CXOToday) |
| MyFitnessPal Meal Scan | Hover the phone camera over the plate | Not fully automatic: users confirm foods, portion and nutrition before saving (BrainStation) |
| Cal AI | Photo, barcode scan, or describe the meal in words | Subscription model; acquired by MyFitnessPal in late 2025, a signal of how valuable photo logging has become (NeuronFeed) |
What it told me
Every leader keeps a human confirmation step, which validated our "Confirm, don't assume" principle. And all of them still need the phone out mid-meal, which is exactly the moment B Glasses removes.
Indirect: manual logging
Classic trackers ask people to search a database, pick a portion and save. Even the photo-first apps market themselves against this: HealthifyMe says food tracking is hard to engage with because users have to type and remember to log (CXOToday).
Effort per meal is the churn driver.
Where B Glasses can win
| Question | Phone AI trackers | Meta glasses | B Glasses Nutrition |
|---|---|---|---|
| Capture | Phone comes out mid-meal | Voice or photo, hands-free | Voice + glasses camera, hands-free |
| Control | User confirms in app | User prompts Meta AI to log | Buddy confirms the dish by voice before saving |
| Uncertainty | User swipes alternatives or edits | Not public | Buddy asks a follow-up ("milk or yogurt?"); AI-generated label on every view |
| Indian plates | Strong in HealthifyMe | US launch first | Designed around Indian dishes and branded snacks |
| Return | Diary and goals | Food log in Meta AI app | Daily Nutrition card and 1-year history; goals and trends are next |
The decision
As a design lead, the decision this drove was simple. We would not compete on recognition accuracy, which belongs to the AI team and improves on its own. We would compete on the moment of capture and on trust, the two places design owns end to end.
07Users and scenarios
One feature, two very different users. Each meets Nutrition through a different moment, and the design had to answer different questions for each.
Regular user
A fast review of the day
- How much have I eaten today, and of what?
- What did I eat on a past date?
- How do I remove a wrong log?
Meets it throughThe Nutrition landing page: daily totals, dates and meal detail
Two user stories, two voice moments
Story 4.1Meal photo breakdown
User
“What's on my plate?”or “How many calories are in this dish?”
Buddy
- Names at least 3 ingredients
- Gives calories, macros and micros
- Offers to log the meal
Story 4.2Log my meal
User
“Hey Buddy, log my meal!”
Buddy
- Confirms the dish by voice
- Replies with its nutrition
- Saves it to the Nutrition card by date
Why it matters
Both stories end with the user in control: 4.1 offers to log, 4.2 confirms before saving. Nothing reaches the Nutrition card without the user's say-so.
08Edge cases and constraints
A voice feature on a camera wearable fails in real-world ways a phone app never sees. Each one had a defined answer before build.
| Situation | How it's handled |
|---|---|
| Ambiguous item | Buddy asks a follow-up, e.g. "Is this milk or yogurt?" |
| Branded snack | Exact brand data first, generic estimate second |
| Dirty lens | Unclear image is treated as an edge case |
| Battery below 10% | Feature unavailable |
| No Bluetooth or weak network | Feature needs both to work |
| Logout, then login | Meal history resyncs |
| Account deleted | Nutrition data is erased |
09User flow
One flow covers both users. A returning-user check skips onboarding; date selection, empty states and deletion all branch off the Nutrition landing page.
Decision points
- Returning user?
- Which CTA tapped?
- Change date?
- Meals logged on date?
- Confirm delete?
10First-time experience: teach the command, show the payoff
A new user has never spoken to their glasses about food. Five screens teach the exact phrase, show what the glasses are doing, and end on where the results will live.
- Intro. A real meal moment and the exact phrase to say, with two paths: Try Now or View Logged Meals.
- How it works. What the glasses and Buddy do after capture: photo, identification, nutrition estimate, and where to view it.
- Listening. Speak after the tone; the LED blinks pink, so the glasses' state matches the screen.
- Analysing. Sets expectations for the wait and points to the dashboard, where the status shows.
- Where insights live. A preview of the landing page, closed with "Got it!"
Intro
How it works
Listening
Analysing
Where insights live
11Nutrition landing page: every state of the day
The landing page is where a regular user spends their time, so every state of the day had to be clearly signalled.
| State | What the user sees |
|---|---|
| Empty | "Nothing on your plate yet" with the voice command in Buddy's gradient |
| Analysing | The new meal appears as a processing row, so nothing blocks the user |
| Logged | Calories, then macros, then the meal list with time and calories |
| Expanded | View More reveals micronutrients; View Less folds them back |
Empty
Analysing
Logged
Expanded
12Meal detail, history and delete
Review any day, correct any mistake.
- Meal detail. Time, dish, calories, macros and micros for one meal; delete sits top right.
- Pick a date. A calendar sheet with Cancel and Apply loads a past day.
- Delete meal. A confirmation sheet explains that the meal's calories and nutrition data will be removed.
Meal detail
Pick a date
Delete meal
Date picker rules. Only dates up to today can be selected, future dates are disabled, and a dot marks days with logged meals. That gives five date states:
- Active
- Meal logged
- Disabled
- Selected
- Today
Date statesdrop the image here
13Key design decisions
Dark, calm interface
Matches the B Glasses app and keeps food photos and numbers in focus
Buddy's voice in gradient
Every voice prompt uses the same purple gradient, so users learn what to say
Progressive disclosure
Calories first, macros next, micros behind View More, so the screen stays glanceable
Visible system status
Listening, Analysing and Logged states mirror what the glasses are doing
Safe deletion
Delete always asks for confirmation and explains that nutrition data is removed
Honest about AI
Each nutrition view says values are AI-generated and may vary in accuracy
The biggest trade-off
Speed against control. Logging should feel instant, but a wrong log damages trust. So Buddy confirms the dish by voice before saving, and the Analysing row keeps the app usable while that happens.
14Requirements to design
Every PRD requirement has a visible answer in the experience.
| PRD requirement | Design response |
|---|---|
| Image trigger: wake word + intent | "Hey Buddy, log my meal" taught in onboarding; Listening screen and pink LED cue |
| Calories, macros and micros | Total intake card, macro row with icons, micronutrient list |
| Clarification dialogue | Buddy asks a follow-up by voice; the exchange is saved as an AI transcript |
| Explicit confirmation before logging | Buddy confirms the dish by voice before the meal is saved |
| Perceived latency under 5 s | Analysing row in Your Meals keeps the app usable while processing |
| Meal logs by date, deletable | Date chevron and calendar sheet; meal detail with delete confirmation |
| Branded snacks | Exact brand data takes priority over a generic estimate |
| Data deletion and resync | Account deletion erases nutrition data; history returns after re-login |
15Measuring success and next steps
What we'll track after launch
- Meals logged per active user per week
- Share of capture sessions that end in a confirmed log
- Days per week a user talks to Buddy about food
- Glasses sessions beyond music and calls
- Meals deleted soon after logging, as an accuracy signal
Next steps, building on Phase 1
- Usability test the first-time flow and voice prompts
- Let users edit a dish or portion, not only delete it
- Feed nutrition data into the upcoming Memory layer
- Add daily goals and weekly trends to the Nutrition card
- Show the clarification transcript inside meal detail
See your food. Know your food.
Sources
- Meta's smart glasses will watch what you eat to track nutrition data, 9to5Google
- Meta Inches Into Health Wearables with New Food Logging Feature, Road to VR
- HealthifyMe introduces AI-powered image recognition for Indian food, TechCrunch
- OWLED Media partners with HealthifyMe on Snap, CXOToday
- MyFitnessPal Introduces AI Photo Recognition, BrainStation
- What is Cal AI?, NeuronFeed