Skip to content
Technology How it works Breeder — Hyperion Burner — Aegis Burner — MetroVolt AI-Native Architecture Magnets Fuel cycle Safety Roadmap
Solutions AI & Data Centers Defense & Government Grid & Baseload Neutron Detection Quantum
Learn Technical Library
Proof Publications Whitepapers Technical Library Open Science & Reproducibility The Honest Gates
Company About / Mission Leadership Environment Health & Safety Investors Careers Press Contact
3D Model
AI Architecture › L4 · Orchestration
L4 · Orchestration

Safety-Envelope Checks

The envelope checker validates that a command keeps the machine inside a state-dependent safe region, not just under individual scalar limits.

THE STACK · click to jumpL7Ecosystem & StrategyL6Experience & VisualizationL5Applications & CopilotsL4OrchestrationL3Twin Modeling & AIL2Data FabricL1Control PlaneL0Foundation▲tlmctl▼L4 · ORCHESTRATIONEvents, workflows, rules, and human routing.1Event Streamingthe backbone2Workflow Enginecampaign procedures3Rules & Safety Boundshard limits4Human-in-the-Loopapproval routing5Schedulerexperiment campaigns6Audit Busfull decision lineageMACHINE TIECoordinates L3 outputs with L5 copilots and human operators.KRONOS FUSION ENERGYAI-NATIVE S.M.A.R.T. GENERATORORCHESTRATIONSHEET 06REV. 2026-08L4 · AI-NATIVE STACK
L4 · Orchestration — its place in the stack (left, click any layer) and its internal components (right). Telemetry rises; control descends.

Beyond scalar limits

Individual quantities can each be legal while their combination is unsafe. The safety-envelope checker evaluates the joint operating point: for the breeder, a plasma current, shape, and pressure combination that risks a vertical displacement or a disruption; for the burner, a plug-to-throat mirror ratio and beta combination that risks losing plug confinement. It is state-dependent, evaluated against the current reconstructed machine state.

Envelope as a region

python
def in_envelope(cmd, state):
    proj = project_state(state, cmd)          # predicted next operating point
    return all([
        g_current(proj)   <= 0,   # each g_i <= 0 defines the safe region
        g_shape(proj)     <= 0,   # negative triangularity delta -0.30 maintained
        g_vertical(proj)  <= 0,   # vertical stability margin preserved
        g_density(proj)   <= 0,   # below empirical density limit fraction
    ])
# projection uses a certified reduced model, NOT the learned twin

Why the checker is not the twin

The envelope projection uses a small, certified, deterministic reduced-order model whose properties are analyzed and bounded, not the full learned digital twin. The twin (L3) may be more accurate on average but is not certified for veto authority. If twin and envelope-checker disagree, the checker wins and the command is rejected. Accuracy is not the same as trustworthiness for a safety veto.

Margins and hysteresis

Placement in the gate

The envelope check runs after schema validation and rules evaluation and before any human gate, so approvers never see a proposal that is already unsafe. See the action-gating pipeline. Every check, its inputs, and its verdict are journalled for replay and audit.

Content reviewed August 2026 · design-and-simulation stage