Lunch decisions in 10 seconds: LunchMate, a decision app
Translated by AIView original
Why I started
At a company without a cafeteria, the same question comes up around 11:30 every day.
"What should we eat today?"
You open a map app and scroll through dozens of places, "whatever" goes around the Slack channel a few times, and if a team dinner gets scheduled, the person handling the bill has to call around to check headcount, budget, and whether there's a private room.
So instead of another restaurant search app, I decided to build an app that just picks where to go right now.
What LunchMate does
- Set your office location, budget, walking distance, and foods you can't eat in one minute, and it recommends 3 nearby restaurant cards for each meal
- Each card has a one-line reason, like "4 min walk, around 9,000 won, Korean soup you liked last week 👍"
- One button for walking directions in Naver Map, and if you still can't decide, "just pick for me"
- Vote with coworkers using an invite link (share via KakaoTalk, live tally, candidates with foods someone can't eat get a warning label)
- Team dinner mode: from headcount, per-person budget, and room/alcohol conditions, it picks 5 candidates and goes straight into a vote
- Next day, answer "How was ○○ yesterday?" with one tap and the next recommendations fit your taste better
The core problem is missing data
Restaurant search runs on the Kakao Local API, but once I started using it, it lacked the info that matters most when choosing lunch.
- No menu prices → From post-meal feedback, I collect the "actual price per person" to build price ranges per restaurant, and show an "estimated" badge when there's none
- No photos either → Instead, the cards are built around category illustrations and big numbers (walking minutes, price, 👍 ratio)
- I can't store place data → The DB only keeps place IDs and the data users leave
- No wait times or discount info → Office workers report "long line" or "lunch special available" with one tap
- And one principle I set: recommendations and ads don't mix. Discount info is only displayed and doesn't feed into the recommendation score.

Next steps
I'm building it with Claude Code, and the first users are Seoul office workers like me. I plan to check whether the recommendations actually lead to visits in a closed beta before opening it up. Feedback welcome from anyone tired of picking lunch and anyone who's been the one handling a team dinner bill!
Comments 0
No comments yet
Be the first to comment.