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 › L0 · Foundation
L0 · Foundation

Data Escalation Tiers

How telemetry is progressively filtered and promoted from raw edge streams to curated foundation datasets along the pipeline.

THE STACK · click to jumpL7Ecosystem & StrategyL6Experience & VisualizationL5Applications & CopilotsL4OrchestrationL3Twin Modeling & AIL2Data FabricL1Control PlaneL0Foundation▲tlmctl▼L0 · FOUNDATIONThe offline compute substrate — multi-physics & batch training.1Cloud HPCelastic burst2Bare-Metal ClusterGPU / CPU3Supercomputingmulti-physics runs4Batch Trainingmodel builds5Simulation FarmGrad-Shafranov · MHD6Object StorecheckpointsMACHINE TIETrains the models that ship UP to L3 — no real-time path to the machine.KRONOS FUSION ENERGYAI-NATIVE S.M.A.R.T. GENERATORFOUNDATIONSHEET 02REV. 2026-08L0 · AI-NATIVE STACK
L0 · Foundation — its place in the stack (left, click any layer) and its internal components (right). Telemetry rises; control descends.

Not all data travels the whole way

The machines produce far more raw telemetry than could ever be stored or learned from in full. Data escalation is the principle that data is progressively filtered and promoted as it moves up the pipeline: everything is seen at the edge, a fraction is archived, and a curated subset becomes training data. Escalation decides what earns the trip to the foundation.

The tiers

At the edge, raw high-rate signals are consumed for control and mostly discarded after acting. The data fabric promotes validated signals and derived features for storage. The archive holds full pulse histories. Curation elevates the most informative of those into versioned training sets. Each tier holds less data but more distilled value.

What earns escalation

Promotion is not uniform. Anomalous or off-normal segments, a disruption precursor in the breeder, a potential-barrier wobble in the burner, are escalated preferentially because they carry the most information for learning. Routine steady-state data is subsampled. Escalation is thus an information-density decision, keeping the rare and surprising while thinning the ordinary.

Escalation is governed by metadata and lineage from the L2 data fabric, so every promoted sample knows where it came from and under what calibration. This is what makes the curated tier trustworthy: it is not a random sample but a governed, traceable selection from the full history.

The tiered design is what makes petabyte-scale operation sustainable for both machines. The foundation never tries to learn from raw firehoses; it learns from a curated distillate, while the full history remains available in the archive for replay whenever a study needs to reach back below the curated tier.

Content reviewed August 2026 · design-and-simulation stage