Howdy,
I’m Anna. I built a page that makes it more fun to search or randomly select a city and learn a few things about it.
I have a family member that loves to read books by authors that lived in a place before she visits the place and so originally I made this for her and it only returned the authors. Then I added the reddit users sentiments and the interesting facts. Except the word interesting is overused so I call them strange facts, although they’re leaning closer to normal facts, but shh.
I’m a product manager by day that loves graphic design and Vercel has made it so I can combine the two in a way that I would have never thought possible without some solid software engineering skills.
There’s a hard rule that “internet voices” must be verbatim lifts from real Reddit posts, not LLM paraphrases. Stage 1 of the pipeline names plausible subreddits; Stage 5 fetches their top posts and runs a verification gate (verbatim string match, sentence-boundary check, content-safety filter, distinct-subreddit rule). If nothing passes, the section ships empty rather than fabricated.
Feedback welcome: especially curious whether the “no paraphrase fallback” rule reads as a feature or a missing piece when you hit a city with sparse coverage. Try a well known city like Lisbon or Buenos Aires for a robust result. Spin the globe to get an obscure city and see what sparse coverage looks like.
Thanks for taking the time to view.