Architecture

AlignTrue is the system of record for AI. The platform captures:

• what AI saw (query artifacts + snapshot semantics)
• why it decided (derived artifacts + lineage)
• how the decision unfolded (trajectories + outcomes)
• what it did (events + receipts + egress logs)

Rings of Trust

AlignTrue uses a ring-based architecture:

• Ring A (Kernel): Non-negotiable core invariants—Events, Commands, Projections. These are “one-way doors” ensuring determinism and auditability.
• Ring B (Platform): Additive contracts built on Ring A, such as Trajectories and Simulation. These provide reasoning context without altering the core state clock.

Planes and clocks

• Truth plane: events → projections (rebuildable, versioned, freshness declared).
• Authority plane: actors → capabilities → approvals/holds → audit.
• Execution plane: commands → handlers → outbox/egress with safety classes.
• State clock: what happened (events/projections).
• Decision clock: how reasoning unfolded (trajectories + outcomes).

Read next

• Invariants (Ring A)
• Envelopes (Ring A)
• Projections (Ring A)
• Trajectories (Ring B)
• Simulation (Ring B)