Behalf

Agents read two public sources per person, write a profile, date each other, and rank who fits. Demo is a finished 25-person run.

What we built, in one breath: Each person is two sources. An agent reads them into needs, hobbies, and interests, then dates the other agents on their behalf. You get profile pages and a ranked fit list.

Run

npm install
npm run dev

Open http://localhost:3000.

• Demo (no typing): /demo — 25 people, profiles, dates, rankings
• Watch agents date: /dates or the featured date on the home page
• Paste your own sources: /new — LinkedIn URL + Instagram URL + the public text from each

How it works

1. Two sources only: a career dossier and a public-life page.
2. The agent analyzes needs, hobbies, interests, qualities, dealbreakers, with quotes.
3. Agents date in a venue drawn from the two lives — dialogue, snag, walk-away.
4. Rankings blend that analysis with what happened at the table.

Technical section — Instagram and LinkedIn

We do not scrape Instagram or LinkedIn. Both platforms prohibit automated collection, and there is no official API for this product.

• Demo of 25: first-party public pages at /s/[person]/linkedin and /s/[person]/instagram (career + public life stand-ins). The agent reads only those texts.
• Live try: you paste the two official URLs and the public text. The analyzer (src/lib/analyze.ts) runs on that text in-process. No fetch of linkedin.com or instagram.com.

Stack: Next.js, TypeScript, React, Tailwind. Matching and dating are deterministic TypeScript agents (analysis + scene generation), not a third-party scrape.

Video notes (≤3 min)

1. Home → a profile (/people/mira-chen) with needs / hobbies / interests.
2. Watch the agents date (/dates/...).
3. That person’s ranking, then /demo for all 25.
4. /new with pasted public text.

License

Demo people are fictional.

AgenticEngineerProject---Dating-App