Lunchmate - Pick your lunch in 10 seconds, a decision app (in beta)
Translated by AIView original
Screenshots
Screenshot 1 of 4
Ten minutes before lunch, I've watched the "what should we eat?" question end in "whatever" on Slack almost every day. Open a map app and dozens of places show up, but picking the one you'll actually go to is hard. Lunchmate is a mobile web app I built to make that decision in 10 seconds. The goal was an app that decides where you go, not one you search for restaurants in.
It's in beta now.
I'm looking for people who'll use it during real lunch breaks until November 10 (Tue) and tell me what they think.
I'll leave the sign-up and setup instructions in the comments.
Who it's for
- Office workers who eat out every day because there's no cafeteria
- Team leads and office admins who have to pick team lunch or dinner spots
- New hires who want to know the restaurants near the office and what lunch actually costs
Today's pick: three options is enough
- Set your office location (or current location), lunch or dinner budget, how far you're willing to walk, and what you can't eat, once.
- Recommendation cards show walking minutes, price per person, and how satisfied other people were, as numbers.
- Each card also gets a one-line reason why this place, like "the pork soup you liked last week" or "a pho place you haven't tried yet."
- Don't like it? Hit reroll. Too lazy for that? "Just pick for me" chooses one place for you.
- It shows fewer places of the same kind as your recent meals, so you don't keep getting the same spot every day.
Deciding with friends: voting instead of group chat debates
- Create a vote room and send the link over KakaoTalk. Coworkers pick on their own phones and the tallies update live.
- When the deadline hits, the top spot wins, and the result screen takes you straight to directions.
- If someone in the group can't eat one of the candidates, it shows "someone can't eat this" without saying who.
Map and directions
- Restaurants within walking distance appear on a map and in a list, and you can filter by category, price, distance, and satisfaction.
- Once you've picked, one tap opens walking directions in Naver Map.
- Restaurant details show price ranges and tags other people left.
- Reports like a closed place or a long line also show up.
Feedback
- The next day when you open the app, it asks "How was the place you went yesterday?" A single 👍 or 👎 is all it takes.
- Actual price per person, tags like "solo meal OK" or "15-min cut," and a one-line review are optional.
- That record feeds into your next recommendations, and the price and satisfaction info becomes useful to other people.
How it was made
I built it solo with Claude Code, from planning through deployment. Serious development started on October 29, and over seven days it piled up about 280 commits.
Most of my time went into building a structure where a person decides and AI carries it out.
To avoid the generic template look that AI-built screens tend to have, I set one strict rule: every new screen starts with 2 to 3 mockups that go in clearly different directions. I capture each mockup, show them side by side in a comparison table, and then stop.
I settled on an ivory background with one olive color. Titles use Malgun Bu-ri serif, and body text uses Pretendard. The linter blocks hardcoded colors, so only design tokens are allowed. I also kept the screen copy in the same tone with a separate style guide and a checking script.
Tech stack
- Next.js (App Router) + TypeScript, Tailwind CSS, deployed on Vercel
- Supabase: Kakao and Google login, Postgres with row-level security (RLS), realtime vote updates (Realtime), deadline handling (pg_cron)
- Naver login is a direct OAuth integration I wrote myself
- Restaurant data from the Kakao Local API, maps from Naver Maps, directions linking to the Naver Map app
- Home screen install (PWA), analytics with PostHog, error monitoring with Sentry
With AI, I could build this much scope solo in a little over a week. But a person still had to decide what to build, which screens fit, and what not to do, all the way through. Writing those decisions down as documents and rules did the most to determine quality. In the beta, I want to see whether people actually use it to decide lunch. If you try it and send feedback, I'd really appreciate it.
- Built with
Comments 2
Translated by AI·View originals
You're invited to the Lunchmate beta. It's an app that picks your lunch for you in 10 seconds. It recommends 3 restaurants near your office, and you can also vote with your coworkers to decide. https://lunchmate-kr.vercel.app/beta Use it to pick lunch from today through Tuesday, November 10. Tell me what's annoying or what you liked by tapping "Send beta feedback" at the very top of the Me tab in the app. Even one line is enough. It's a test build, so the screens and features may change. Your current location stays on your phone and isn't sent to a server.
[1-minute install guide] How to put an icon on your home screen without the app store ① Open the link and pick "Open in external browser". ② On iPhone, tap "Add to Home Screen". The steps are a bit different in Safari and Chrome. On Android, tap the "Install" button on the screen. If there's no button, pick "Add to Home screen" from the Chrome menu (the three dots). ③ Open the Launchmate icon that appeared on your home screen and log in.