How the MT edition is designed to work.
JustineAI™ MT runs the same compositional fabric as the in-market PI edition — Eve-Legal F5/reasoner, one Justine supervising stage-specialized sub-agents in a single 10M-token context. What changes is the calibration. Each deep-dive below describes a mechanism this edition is built to carry once it opens. The architecture is real today; the MT calibration is on the roadmap.
- 01Architecture
One supervisor over thousands of plaintiff sub-agents
When the Mass Tort edition opens, one Justine will hold an entire inventory in a single reasoning context — every plaintiff at once.
Read deeper → - 02Bellwether
Bellwether selection scored against the CMO
When CMO 1 sets the trial-selection protocol, Justine is designed to score the inventory and defend a representative slate to the special master.
Read deeper → - 03Settlement
Modeling the global settlement matrix
Justine is designed to model an inventory deal as a tiered grid before the committee agrees to it — every tier definition costed in advance.
Read deeper → - 04Liens
Lien resolution at pool scale
Before a single check is cut, Justine is designed to compute every plaintiff’s net — and surface the claimants who net to zero in time to fix it.
Read deeper → - 05Discovery
Common-benefit discovery, reasoned once
Justine is designed to read the corporate document depository in one context and find the notice and causation proof the whole pool shares.
Read deeper →