I built AI character chat that remembers the way people do
Another AI character chat wrapper with a clean UI but no technical differentiation.

Character AI meets Foursquare, but the location layer is more aesthetic than functional.
People interested in AI companions and location-based experiences
Character AI · Replika · Pi
I’ve always liked the idea of talking to strangers in places like cafés, bars, airports - just to hear their stories. But in reality, people are busy, guarded, or just not in the mood.
So I started wondering: what if you could walk into a place and talk to an AI character instead?
I built a map-based experience where you can open a city, step into real cafés and bars, and meet AI characters there to talk to. It’s somewhere between Foursquare and Character AI.
Some example characters
- A staff engineer rewriting the same system for the third time, late-night café in San Francisco
- A founder who just shut down their startup after 4 years, alone at a bar in New York
- A 19-year-old open-source dev, coding in a dim café in Berlin
- A crypto trader who made and lost a fortune twice, at a bar in Dubai
- A strip club dancer saving up to leave, between shifts in Las Vegas
Some of the conversations are surprisingly engaging.
Current state: 5 cities, 100+ characters
I’m mainly testing:
does this feel interesting enough to explore?
Would love to know what you felt when you tried it.
Another AI character chat wrapper with a clean UI but no technical differentiation.
Native macOS menu bar for live F1 timing when browser tabs eat your RAM.
Fully local transcription with auto meeting detection beats cloud services on privacy.
Live-typing chat is novel, but Discord and Slack already dominate this crowded category.
LLM-extracted lore plotted on tile-zoomed map built entirely on economy WiFi.
Lives in the menu bar, uses EventKit and local-only logic, and auto-opens Zoom/Meet/Teams links at a configurable time — a small, practical win for anyone who scrambles for links. Switching from Electron to Tauri to shave bundle size is a nice engineering choice, but the idea itself is already served by existing menu-bar tools; what'll sell Galopen is polish around edge cases (multiple links in one invite, recurring/timezone quirks) and a clear UX decision on confirmation vs full automation.