Last month I went to Indonesia, and was eating at a local place with friends who did not eat raw fish&green onions. We tried ChatGPT (of course) it worked, except it wasnt a smooth experience: taking too long, and difficult for the staff to provide proper feedback quickly.Then I realized I only needed to tell the waiter what we won't eat, and ask for recommendations, that's all. So I made MealWords (recommend to try on your phone):1. You pick allergies, dietary and preferences in your language (peanuts, no cilantro, not spicy etc) and it shows the card in the staff's language.2. The waiter taps anything they can't do on the card, or just confirms. You read the answer in your own language.On translations: nothing is machine-translated (well, llm-assisted for research but not mechanical translations). For allergens&terms I used all official guidance I could find (FDA, EFSA, Japan etc all official governments and local public health orgs) to make sure I used not only accurate but also understandable terms. Then I worked with native speakers (aka my friends all over the world) to proofread it as plain kitchen language, NOT medical wording (if you speak one of the languages and something feels odd, I'd really like to know!) Also, wordings were carefully chosen to meet local culture (for example if the target language is Chinese, there are extra explanations about peanut allergy when chosen, and what local sauces should be avoided in Chinese, where the general are not familiar with the concept enough compared to other places like Japan)Technical bits: the website is static, no account, card and choices stay in the browser (page analytics etc are described on the privacy page as they are but no privacy is collected) Comments URL: https://news.ycombinator.com/item?id=50286113 Points: 2

Comments: 0