The platform
How the Aplark platform composes.
One product exists today. The architecture below is how the rest is intended to fit around it, and the shape of that architecture is itself a commitment.
Intelligence Core
- Perception
- Knowledge
- Reasoning
- Simulation
- Risk
- Autonomy
- Assurance
Aplark Core
The connective layer
Aplark Core is the evidence substrate the rest of the platform composes around: content-addressed artefact storage, an append-only tamper-evident ledger, environment fingerprinting, canonical serialisation and the signing chain. It is not a product you buy separately — it is the part that makes anything above it citable.
Its design constraint is unusual and deliberate. Nothing in Aplark Core produces a model, a dataset or a rendered scene. It only records, orders and proves. Every capability added to the left of the ledger dilutes the one property that is genuinely defensible: being the authoritative record of what happened.
The trust boundary
No product receives privileged status.
Aplark Assure will never produce the thing it evaluates. Any Aplark tool that generates data, renders simulation, trains, tunes or optimises a model sits outside Assure’s trust boundary as a separate product, and its output enters Assure only as a declared, provenance-tracked input.
No sibling product receives privileged status. Data originating from an Aplark generation tool carries the same declared source type, the same default intra-cluster correlation, the same mandatory sim-to-real-gap defeater and the same provenance requirements as any third-party source. There is no code path, configuration flag or licence tier that relaxes this because the data came from us.
A vendor that both generates the test data and adjudicates the sim-to-real gap holds a conflict that a competent certifier identifies in the first meeting. Stating the boundary publicly, and building to it, converts the weakest question in the room into the strongest answer.