IDA and JUDGE
Coherence and legitimacy pressures for understanding a situation without losing the question of whether a proposed response is justified.
Multi-agent intelligence
Plural intelligence for accountable systems.
The IDA family is an internal separation-of-powers experiment: distinct student bodies carry different legitimate pressures so interpretation can preserve useful tension instead of flattening every question into one voice.
Many minds deliberate. One voice speaks.
IDA can interpret, challenge, explain, and escalate. It does not own authority. Governed context and attributable human or institutional review remain the action boundary.
Family design
Different pressures, shared responsibility.
Each student role contributes a distinct signal to the family state. The point is not to create identical assistants; it is to make disagreement, evidence, caution, memory, and trajectory legible before convergence.
SENTINEL and SHADE
Safety, exposure-control, and boundary signals that keep risk and concealment from disappearing inside a fluent answer.
ECHO, ATLAS, VECTOR, and more
Memory, trajectory, repair, urgency, equilibrium, and action pressures help the family carry context across time.
Research and training
Model bodies with a visible lineage.
The public research trail includes the compact IDA Edge body, the larger IDA AI body, the IDA MoE mixture-of-experts body, and the IDA Lattice tokenizer. Training, evaluation, and promotion remain separate decisions; a checkpoint is not a capability verdict.