The Honest Gates
The architecture states plainly what is designed, what is simulated, and what is not yet proven — no capability is claimed ahead of the physics or the schedule.
What the honest gates are
The honest gates are the explicit caveats the architecture carries so that its claims never outrun its evidence. They separate what is built from what is designed, what is measured from what is simulated, and what is proven from what is targeted.
The gates
- Design and simulation, not hardware. Both machines are studies today; the control stack is validated against physics codes, not an operating reactor.
- No net-gain claim before FOAK. Construction begins Q2 2027; first-of-a-kind first tritium is targeted around 2030; no hardware net-gain is claimed before then.
- Frozen canon only. Quoted numbers — Q_sci 3.076, 85.0 MW, 9.66 MA, 16.84 T, delta -0.30, 26.49 T plug, 5.44% neutron fraction — are the design point, stated with their conditions.
- The tritium breeding ratio is a lever, studied across 1.1, 1.5, and 1.8, not a settled result.
- D-3He is low-neutron, not aneutronic; p-11B is a gated research endpoint only.
Why gates belong in the architecture
An AI-native plant is persuasive precisely because it produces confident predictions. That makes disciplined honesty more important, not less. The gates are wired into the governance layer so that a copilot or dashboard cannot present a simulated result as a measured one, and cannot promise a capability the schedule has not reached.
Gates and the learning loop
The gates also frame the retraining loop. Until FOAK produces real data, the twin is trained on simulation and will carry simulation's biases. Pink feedback and batch retraining are how those biases are corrected once real operation begins — the gates make clear that this correction is still ahead, not behind.
The gates are enforced by data lineage and governance and reflected across every page's framing.