What 'AI-Native' Means for a Fusion Plant

AI-native means the intelligence is not a bolt-on analytics layer but the control fabric itself — designed in from the plant's first block diagram.

Native, not bolted on

Most industrial AI is retrofit: a plant is designed conventionally, then instrumented and analyzed. Kronos is AI-native, meaning the architecture assumes machine learning, the digital twin, and autonomous control from the start. The diagnostics, the data fabric, and the control plane are specified together so that the machine is observable and controllable by design.

What changes when AI is native

• Diagnostics are chosen for what models need to infer, not only for what humans read.
• The data fabric is built RAG-ready and feature-store-first, not as an afterthought log.
• Control is model-predictive over a live twin, not a fixed set of PID loops alone.
• The plant improves continuously through a retraining loop wired in from L0.

Why fusion needs it

A fusion plasma is a coupled, nonlinear, high-dimensional system that evolves faster than a human can react. The breeder must avoid disruptions while holding a negative-triangularity equilibrium; the burner must sustain an ambipolar potential against end-plug density fluctuations. Neither is a set-and-forget process. AI-native control is what makes real-time reasoning over these systems tractable.

Native does not mean unbounded

AI-native does not mean AI-in-charge. The most consequential actions — quench protection, vacuum interlocks — are deterministic and AI-independent. Being AI-native means intelligence is designed into the observability and decision layers, while the safety floor remains purely physical. The design philosophy makes that split explicit.

Contrast with retrofit analytics, and see how the split is enforced, in determinism versus intelligence.