L6 Experience: The Human Interface to Two Fusion Machines

Layer 6 is where operators see, trust, and act on everything the lower layers compute — for the breeder (Hyperion) and the burner (Aegis / MetroVolt).

What Layer 6 is responsible for

Layer 6 (L6) is the human-facing surface of the KRONOS AI-Native stack. It consumes state estimates, forecasts, confidence scores and provenance from L3 (twin modeling), orchestration status from L4, and copilot reasoning from L5, and renders them into control-room displays, plant-floor SCADA/HMI, mobile engineering views, and alerting workflows. L6 owns no physics and no actuation authority: the L1 Control Plane and its hardware failsafes remain the sole real-time authority. L6's job is situational awareness and decision support — making the machine legible to a human on a timescale humans can act on (roughly 0.5 s to minutes).

Because these are design-and-simulation machines not yet built (breeder construction begins Q2 2027, first-of-a-kind first tritium ~2030), L6 today is being co-designed against the digital twin. Every display is wired to the twin's simulated state first, so operator workflows are exercised long before first plasma. Commissioning then swaps the twin's synthetic telemetry for live diagnostics with no change to the operator's mental model.

The two machines behind one interface

The breeder (Hyperion) is a D-T spherical tokamak: Q_sci 3.076, 85.0 MW fusion power, 9.66 MA plasma current, 16.84 T peak field, negative triangularity delta -0.30, tritium breeding ratio treated as a design lever (1.1 / 1.5 / 1.8). Its operator views center on equilibrium, the ELM-free shape, and disruption avoidance. The burner (Aegis / MetroVolt) is a D-3He tandem-mirror generator: 26.49 T plug, 17 T throat, 5.44% neutron fraction, direct energy conversion. Its views center on end-plug density, the ambipolar potential, and DEC grid modulation.

• In: L3 state/forecast/confidence, L4 job status, L5 copilot narratives
• Core: 3D twin overlays, SCADA/HMI, dashboards, alerting, incident replay
• Out: operator intents (advisory), acknowledgements, override requests
• Never: direct hard-real-time actuation — that stays in L1

The pages in this category specify how the twin is drawn over the plant, how uncertainty and model provenance are made visible, how alarms are rationalized to prevent floods, and how the control room is laid out to keep cognitive load survivable during a disruption countdown or a plug-density excursion. See the 3D twin overlay and uncertainty visualization to start.