Matterforma is built on explicit objects that allow orchestration to be repeatable. Every decision can be traced to a record.
An Artifact is a versioned Molebyte with immutable identity, declared constraints, metadata, and provenance links. It is the unit that agents generate, humans review, and policy evaluates. Artifact immutability ensures every future run references a known logic boundary rather than mutable source. This lets teams compare outcomes over time without ambiguity and guarantees that policy decisions map to an exact executable object.
A Run is a scheduled execution instance bound to one Artifact version, one profile, one policy context, and one pool decision. Runs are the operational bridge between intent and execution. They include lifecycle states, timing metadata, and failure semantics so operators can reason about throughput, confidence, and governance. Because runs are explicit objects, automation can be monitored, replayed, escalated, or blocked with clear attribution.
A Signal is the structured outcome of a Run, including distributions, confidence bounds, state deltas, and economics. Signals replace unstructured logs in the Material Computing model. They are designed for machine interpretation and human review at the same time. Signals become meaningful when compared against baselines, which is why every signal retains linkage back to run context and artifact identity for repeatable analysis.
Policy defines the constraints under which automation can operate: risk classes, approvals, cost ceilings, profile restrictions, and pool eligibility. Policy is enforced by Immunity and evaluated before execution. In orchestration, policy is not documentation; it is executable governance. This ensures that agent behavior is bounded by organizational intent and that exceptions require explicit, auditable intervention instead of implicit shortcuts.
Diff is the comparison result between a Signal and one or more baselines: previous runs, expected distributions, reference artifacts, or threshold policies. Diff translates raw outcomes into actionable variance. It captures what changed and how materially it changed. Diff is central to long-running automation because programmable logic can still drift when environments or constraints evolve.
Judgment is the decision outcome produced after policy evaluation and diff analysis. Typical outcomes are approve, rerun, route, escalate, or stop. Judgment is where control-plane reasoning becomes operational action. By making judgment explicit and recordable, Matterforma enables safe automation: agents can act quickly, and humans can inspect exactly why each action was taken.
Route is the next execution choice selected by judgment. It determines whether a workload should remain on silicon, move to Material Cloud execution, switch pools, change profile depth, or pause for review. Route closes the loop and starts the next cycle. Route decisions combine confidence, economics, and governance, making orchestration adaptive without losing policy boundaries.
