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

How it works · Chapter 7

A memory, not a library

How Hamsa gets better with use without retraining, and how it can grow past one newsroom.

Every verdict a checker publishes, true or false, is fingerprinted into Hamsa's archive, and so is any original footage a checker finds by hand. The next copy of that footage, even re-encoded, cropped or re-captioned, is recognised without anyone searching. An exact repeat of a link is answered at once with the published verdict.

It does not need to know everything

A fair worry: the archive starts small, and a newsroom would need years of footage to make it complete. But it never has to be complete. Footage nobody has checked goes to a person anyway. The archive only has to know what checkers have already decided, and that grows with every verdict.

Growing past one desk

The archive stores fingerprints, about 100 KB per keyframe, never the videos. That makes them easy to share. If newsrooms exchanged fingerprints with their label and a link to the verdict, a fake debunked at one desk would be caught at every other. Platforms already fight known harmful media this way, sharing hashes rather than content. Each desk would choose whose verdicts it trusts.