The Hamsa Book · contents

The story

  1. 1The flood, and the videos that came after
  2. 2A triage desk, not an oracle
  3. 3One post, from a phone to a verdict

How it works

  1. 4Seeing footage before: three fingerprints
  2. 5Finding the evidence, in two languages
  3. 6Models propose, rules decide
  4. 7A memory, not a library
  5. 8Under the hood: one laptop, open models

The proof

  1. 9The numbers, and what went wrong
  2. 10What Hamsa does not do yet
← Back to Hamsa

The proof · Chapter 10

What Hamsa does not do yet

Where it stops today, and what comes next, in the order the evidence points to.

  • Recognise footage nobody has checked. New footage has nothing to match. It is judged on news, fact-checks and the web, and always reaches a person first.
  • Reverse-search every image automatically. No free API for that exists. Checkers keep one-click Google Lens, Yandex and TinEye, and Hamsa remembers what they find.
  • Open private or deleted posts. The sender is asked for the photo or video instead.
  • Match every repost style. A video placed over a blurred copy of itself, common on TikTok, is not matched yet.
  • Work beyond disasters. Only the news filter, a few prompt lines and one rule know about disasters. But each new topic, such as elections or health, needs its own tested examples before any quick check is trusted.
  • Run in a newsroom, yet. No fact-checking desk has used Hamsa in daily work so far.

What comes next

  1. Send "not checkable" posts that carry disaster footage to a person instead of dismissing them.
  2. Grow the held-out test set, and keep scoring real posts as checkers publish.
  3. Match videos placed over blurred copies of themselves.
  4. Share fingerprints between newsrooms.
  5. Read Content Credentials (C2PA) where a camera or editor has signed the media.
  6. A WhatsApp tipline, more languages (Maithili, Bhojpuri), and topics beyond disasters.