Find your person through the posts you both love
An agent that reads your public social footprint, maps your affinities, and introduces you to the one person whose quiet patterns match yours.
The problem
Dating apps reduce people to a few photos and a bio. You swipe through strangers based on seconds of attention. The deeper signals (what you save, what you share, the tone you use, the ideas you consistently return to) are invisible to the interface but obvious to anyone who watched you from a distance over time.
What AI makes possible now
Models now read unstructured text, images, and audio at scale with real comprehension. An agent can ingest years of your public posts, saved items, and comment tone across platforms, then build a dense affinity fingerprint. It does the same for others who opt in and surfaces the nearest match, someone whose pattern of curiosity, humour, and care aligns with yours. Voice and image generation let the first conversation feel less cold without either person writing a single line of copy.
How it works
- You connect the platforms you want the agent to read: public tweets, Reddit saves, Spotify playlists, Substack highlights, Instagram saved posts.
- The agent processes your footprint into a multi-dimensional affinity vector. Topics, emotional tone, vocabulary, cadence of posting, what you linger on versus what you scroll past.
- Weekly, the agent compares your vector against the opted-in pool and returns one introduction. The match is presented with a plain paragraph explaining why the connection is unlikely to be noise.
- You accept or decline. If both accept, the agent writes a short, warm opening message in your voice and generates a five-minute audio clip each of you can listen to before replying.
The first thirty days
Ship a single-platform reader that ingests a user's public Twitter timeline and saved items, builds a basic topic and tone vector, and returns one match per week with a concise explanation. No chat. No audio. Just a name, a reason, and a one-click opt-in to reveal profiles to each other.
How it earns
Free during seeding to build the opted-in pool. Then a flat monthly subscription for weekly introductions. A higher tier adds cross-platform reading, voice notes, and the agent-written opening message. The value is a quieter, higher-signal introduction than any swipe app can offer.
Why now
Cheap inference means reading a person's entire public history costs pennies, not dollars. Multimodal models handle text, images, and audio in one pass. And there is enough ambient fatigue with performative dating profiles that a silent, observation-based alternative feels like relief, not surveillance.
First customers
People over 28 who have left dating apps twice and feel their real self lives in what they share, not in what they pose. Start in communities where public curation is already intimate: Substack writers, Letterboxd users, Ravelry knitters, niche Reddit contributors.
The hard part
The product must earn deep trust before it reads anything. If the match explanation feels generic or the pool is too small to return a genuine connection, people will dismiss it as another black box. Reputation collapses on the first mismatch that feels creepy or random.
Build this with us
The agent's reading style and the match explanation paragraph need careful tuning to feel perceptive without overreaching. Nexibeo can build and run the full ingestion, vectorisation, and weekly matching pipeline as a managed AI product on one fixed fee. If you know the community where this should start, we should work together.