Skip to main content

​
Overview

MetricValue
DateMarch 21, 2026
DomainOncology — ovarian cancer vs endometriosis
Agents24
Packets1,502
Parse validity100%
Compression ratio10.41x
Fidelity scoreF = 97.6

​
The Question

Patient 7291, female, age 47. CA-125 at 285 U/mL (8.1x elevation). Complex cystic mass on ultrasound. History of endometriosis. Does she have ovarian cancer or endometriosis?

​
The Deliberation

24 agents — oncologists, pathologists, radiologists, surgeons, patient advocates, clinical guidelines databases — debated across 12 rounds using only AXL packets. Key moments:
  • Round 3: Oncologist infers malignancy (INF.82) from CA-125 + imaging
  • Round 5: Pathologist contradicts (CON.65) citing endometriosis history
  • Round 7: Oncologist yields (YLD.55) — changes from malignant to uncertain
  • Round 10: Merge (MRG.70) — both viable, biopsy indicated
  • Round 12: Prediction convergence at 60% malignancy probability

​
The Compression Proof

The same deliberation in English consumed 15,647 characters. In AXL: 1,502 characters across 1,502 packets. Ratio: 10.41x All 24 agents. All clinical data points (CA-125, HE4, ROMA score, imaging findings). All confidence levels. All belief changes. Compressed into one-line packets.

​
3D Visualization

This experiment has been rendered as both orbital and cockpit-paradigm 3D visualizations using Blender on an RTX 5090, showing agent positions, connection curves, and belief trajectory evolution in spatial form.