How it works, and what it does not do

Who Would Love is a demo built for the Qloo Agentic Hackathon (2026). The source code is open: github.com/crjstobal/who-would-love.

Methodology

  1. Read. Firecrawl fetches your page (or your home page plus up to four key pages). Claude reads it and describes the product, then splits the likely audience into three or four segments, each backed by a quote from your page.
  2. Match. For each segment, Claude proposes a few well-known books, shows, brands, artists or people that capture its taste. We look each one up in Qloo and Claude checks every match; anything Qloo does not have is dropped and listed, never guessed.
  3. Ask the taste graph. For each segment, Qloo returns what else these people love (overall, in a niche popularity band, and within the main market), how the audience leans by age and gender, and where the taste is strongest.
  4. Find the doors. Claude searches the web for communities, newsletters and podcasts that fit each segment, guided by the Qloo results. Every link is checked before it is shown.
  5. Write. Claude writes the advice. It picks Qloo results by id, and the numbers you see are copied straight from Qloo, so the model cannot invent a score.

Every block in the report is labelled as Qloo data, web research or our interpretation, and you can filter by label.

Limits

Privacy

We do not ask for your name or email and there are no accounts. We store the link you submit, the finished report, and a salted hash of your IP address for rate limiting. Only public content from your page and product-related entities are sent to our providers (Firecrawl, Anthropic, Qloo). No personal data is ever sent to Qloo. Reports are reachable by anyone who has their link.

Terms

This is a free, experimental demo with a small daily limit. Reports are generated by AI and taste data and may contain mistakes; use them as a starting point for your own judgement. Do not submit pages you do not have the right to analyse. The service may change or stop at any time.

Contact: hi@javi.link