关于
关于
what to eat?
A meal-generator app to save the day
the problem: what's lunch?
My partner and I love food.
He resides 12 hours ahead of me in Singapore, and is always thinking about what to eat, especially where he can grab a quick power bite around his office during lunch break. "What should I eat for lunch today?" seems like a tie that connects—except that I always give the "wrong" answer. After years of failing as an understanding partner, I've decided to design a tool to save the day.
Introducing: The Food Wheel!
the challenge: what are the options?
Analyzing Food Pattern
Understanding what he likes to eat is not the hard part; the hard part is making a good guess.
I downloaded the image files he sent from our WhatsApp chat history for the last year; the result was an album of 278 images.
Cataloguing:
Several factors that impact his food choices are:
-
time of the day - breakfast, lunch, or dinner
-
time of the year - work-day quick lunch vs holiday family gatherings, etc.
I utilize two different LLMs to cross-reference - ChatGPT and Claude - to analyze the images, providing me with two text-based spreadsheet catalogues, and labeling the food as breakfast, lunch, or dinner based on the time sent, along with the date.
Data Validation:
Based on this information, I used the two chatbots to create two sets of "food reports", including data visualization of the most common dish eaten, food patterns depending on time of the year, etc. After cross-comparing the two sets of results, I chose to continue with the one Claude created, and shared the report with my "client" for feedback.
The analysis reports reveal some overlooked patterns:
-
Toppings: Pork is a consistent protein choice at various fast food places offering "make your own bowl" options, whether grilled, in soup, or atop noodles and rice.
-
Palette Change: Food options evolve; for instance, spicy Chinese cuisine has recently become a regular part of the diet since the fall.
-
Seasonal Influence: habits fluctuate throughout the year. During holidays like Lunar New Year, there's a shift from single-serving fast food to larger family gatherings away from the office.
the design: prototyping and finalizing
I envisioned a simple, straightforward product that serves only one job - give the user an option, and preferably hit the jackpot. If not, at least help the user decide what they do NOT want to eat.
select preferences
breakfast / lunch / dinner / surprise
generate
results
food for the day
the earlier prototype i build with claude code
Inspired by retro games, I created the design with a lighter pastel color palette that reflects an 80s retro aesthetic, and each dish features a retro pixel icon. Since many dishes allow for "choose your own toppings" adventures, the tool will also provide potential toppings for inspiration.
I also added some snacks and surprises within the options, noted as the "chaos pick". They are less healthy, but maybe suggest a sign for a needed "guilt pleasure".
the result: not sure if it saved the day
After the tool was circulated among several users (a.k.a. some colleagues), it gained some fun and giggles in his office. I've received messages such as "I followed the call from the wheel today!" that felt very good on my end as the product owner.
However, it did not get used for long. At the end of the day, there are a lot more factors that cannot be simulated by technology, and what's more, it simply cannot replace human conversation tied to the text message -
"What should I eat for lunch today?"
