How We Prove It

Most demos show a system working.
Ours tries to break it.

Any platform looks good on a clean network with a cooperative dataset. Intelligence work does not happen there. It happens across dispersed sites, over links that fail, with analysts who disagree.

So our demonstration is built the opposite way round. It sets up the conditions that destroy lesser systems — a site going completely dark, two analysts writing contradictory assessments at the same moment, a link that drops mid-sync — and then proves, with evidence printed on screen, that nothing was lost.


Twelve acts. Each one asserts a fact and prints PASS or FAIL.

A single command rebuilds three independent headquarters from empty — three separate databases, three separate processes, mutually authenticated links — and runs the demonstration against them in about ninety seconds.

  • Three independent sites hold a provably identical picture
  • A collection at one headquarters converges to the whole federation
  • Cross-site references are translated, never silently mis-linked
  • Concurrent edits at two sites both survive
  • Submarine mode — a site is killed mid-patrol, both sides keep working, and it rejoins with lossless catch-up in both directions
  • Priority tiers — the analyst’s thinking returns before the bulk data
  • Sensitive fields cross the federation as ciphertext
  • Every change carries origin-signed provenance
  • Investigations and notes federate with their references intact
  • A true conflict — two partitioned sites write contradictory assessments to the same field. One value wins everywhere, and the loser is preserved for analyst review
  • Every merge class is exercised
  • A regression guard against a real bug that once stranded data silently

It can fail. That is the point.

The demonstration is not a video and not a script. It asserts each fact against three live databases, and if something is broken it says so. Nothing we claim is anything we cannot show.

Proof, not “probably”

Every site computes a cryptographic digest of its own database. Three separate systems showing the same digest is not a claim that they are in sync — it is a demonstration that the pictures are identical.


The shared picture builds across three organisations. After each step, all three databases agree exactly.

The warship goes dark, and its analysts keep working. Its own record changes with them.