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 › Real-Time Control & Safety
Real-Time Control & Safety

Voting and Redundancy Architecture

Critical safety signals are triplicated and voted so that a single failed channel is outvoted and flagged, not obeyed.

STRATEGY / SLOW ▲ ▼ MICROSECOND REAL-TIMEL7Ecosystem & Strategytelemetry ▲ control ▼open ▸L6Experience & Visualizationtelemetry ▲ control ▼open ▸L5Applications & Copilotstelemetry ▲ control ▼open ▸L4Orchestrationtelemetry ▲ control ▼open ▸L3Twin Modeling & AItelemetry ▲ control ▼open ▸L2Data Fabrictelemetry ▲ control ▼open ▸L1Control Planetelemetry ▲ control ▼open ▸L0Foundationtelemetry ▲ control ▼open ▸PHYSICAL S.M.A.R.T. GENERATOR PLANTBREEDER · HYPERION1R0 1.2 m · A 2.5 · 16.84 T · δ −0.30BURNER · TANDEM MIRROR2317 T throat · 26.49 T plug · fₙ 5.44% · DEC1 center stack + plasma · 2 high-field plug · 3 expander → direct converterCOLOR GRAMMAR strategy AI-workflow infra/data models reactor/DECLINE SEMANTICStelemetry (µs)controlKRONOS FUSION ENERGYAI-NATIVE S.M.A.R.T. GENERATORMASTER BLUEPRINTSHEET 01REV. 2026-08L0-L7 · 2 MACHINES
The AI-Native S.M.A.R.T. Generator Master Blueprint — eight layers (L0→L7), one control stack, wired to both machines. Telemetry rises in microseconds; control descends the same path.

Two-out-of-three, and why

A single sensor or channel can fail in a way that reads plausible but wrong. Kronos triplicates the most critical safety signals and votes them two-out-of-three (2oo3). A single deviating channel is outvoted, the machine keeps running on the agreeing pair, and the odd channel is flagged for maintenance. Two channels must fail the same way at the same time to defeat the vote, which is far less likely than one.

Ch ACh BCh CVotedFault flag
1111none
1101C
1011B
0111A
1000A
0000none
python
def vote_2oo3(a, b, c):
    voted = (a and b) or (a and c) or (b and c)
    disagree = [name for name, v in (('A',a),('B',b),('C',c)) if v != voted]
    return voted, disagree            # value + which channel to flag

Degrading gracefully

When one channel is flagged out, the survivors form a 1oo2 pair. The design chooses whether the remaining pair votes conservatively (either can trip) or agreeing (both must confirm) based on whether the function is trip-critical or availability-critical. Safety-critical trips favor 1oo2 trip-if-either; the trade is a slightly higher spurious-trip rate, accepted deliberately.

The choice between 1oo2 trip-if-either and 2oo2 confirm-both after a channel is lost is made per function and recorded, because it trades spurious-trip rate against missed-trip rate and only the hazard analysis knows which way to lean. Safety-critical protections accept more spurious trips to guarantee real ones; availability-critical but non-hazardous functions lean the other way. Voting logic itself is kept simple enough to verify by inspection, since a clever voter is a common-cause hazard.

Redundancy defends against random channel failure but not against a common cause that takes all three at once — a shared power rail, a shared calibration error, a shared firmware bug. That gap is closed by diversity, covered in common-cause failure mitigation. Voting composes with watchdog supervision, which itself votes on liveness.

Content reviewed August 2026 · design-and-simulation stage