Online dating today is noisy, time‑consuming, and too often unsafe. Endless swipes, low‑effort messages, recycled prompts, and fake or manipulated profiles erode trust. I wasn’t interested in building yet another “get more matches” gimmick. I wanted something that made the core experience safer, clearer, and higher quality for people who actually care about meaningful conversations.
The project started from one recurring frustration: you can’t easily tell who is real, who is genuinely engaged, and whether a chat is moving in a healthy direction. Most tooling in this space optimizes for dopamine loops, not trust. That felt backwards.
1. Catfish Detection: Lightweight pattern checks, linguistic signals, and metadata cues help surface profiles or message patterns that look synthetic, low‑effort, or risky before you invest time or emotion.
2. Interest Analysis: The assistant looks at conversational turns to infer engagement signals: curiosity, reciprocity, topic depth, emotional mirroring so you can distinguish polite replies from genuine interest.
3. Conversation Quality: Clairity, balance, pacing, escalation readiness (are you both moving toward meaningful next steps?) all surfaced as a simple qualitative snapshot instead of abstract scores you can’t interpret.
Yes, there is a reply helper but it’s deliberately constrained. It’s there to unstick you, not to ghost‑write your personality. The focus is insight first, generation second.
The app runs locally. No trackers. No hidden API calls to some opaque backend. You bring your own AI provider key; requests go directly there. The code is open source so claims about privacy aren’t marketing they’re inspectable.
Upcoming work: richer anomaly heuristics for catfish detection, optional on‑device model scoring, nuanced emotional trajectory timelines, and a privacy‑preserving way to share anonymized red‑flag patterns (opt‑in only).
If this resonates, try it, inspect the code, and suggest sharper heuristics. The goal isn’t to gamify dating it’s to reduce wasted emotional bandwidth and raise the quality of interactions.