The throttle is an interpreter, never a source of truth.

The hierarchy

Authority descends. Evidence ascends. The throttle sits near the bottom.

The full stack runs from the journal at the top — raw recorded reality — down through the weekly summary, the EV and compliance engine, the structural diagnostics, and the gate-and-brake state engine, before reaching the throttle, which hands directly to execution. Each layer consumes the layer above and produces a narrower, more operational artifact. By the time the chain reaches the throttle, every question of truth has been answered: what happened, what it means, whether the structure is healthy, what capital state the account is in. The throttle receives settled facts and owes the stack exactly one service in return: translate those facts into a next-cycle deployment decision that respects every constraint they impose.

What the throttle may do

Modulate inside boundaries. Nothing else.

The panel's legitimate powers are real but bounded: it selects the final tier within the gate's ceiling, prices the cycle pool from the gate's budget row, resizes fresh deployment around open exposure through Smart Capacity, applies the graded structural suppressions to the candidate tier, and records every decision in the cycle log. What it may never do is equally explicit: it cannot override gate contraction geometry, cannot soften the drawdown doctrine, cannot skip a brake tier, cannot unlock a System Lock, and cannot second-guess a structural verdict. The distinction is between modulating deployment inside allowed boundaries and moving the boundaries — the first is the throttle's entire job, the second is a constitutional violation regardless of how good the reasoning looks that day.

Why the boundary exists

The layer closest to execution faces the strongest temptation.

The throttle is where the system touches money, which makes it the layer under the most pressure to 'be reasonable.' Every argument for bending a rule arrives here first, dressed in operational language: the setup is exceptional, the diagnostics are lagging, the gate hasn't caught up to the recovery. If the throttle could adjudicate those arguments — if it were permitted to weigh structural truth rather than receive it — then every upstream layer would become advisory, because their outputs could be re-litigated at the point of deployment. The separation of diagnosis from deployment is what makes the diagnosis binding. The SDE diagnoses structural condition; the calculated diagnostics generate the overlays and vetoes; the throttle converts vetoes into behavior. Three responsibilities, three owners, zero overlap — and the architecture stays trustworthy precisely because the most tempted layer is the least empowered.

• SDE: diagnoses structural condition. Never sizes anything.
• Calculated diagnostics: generate overlays and vetoes. Never execute.
• Throttle: converts vetoes and states into deployment. Never diagnoses.

The interpreter test

Every throttle output must be reproducible from upstream state.

There's a concrete audit that enforces the interpreter doctrine: given the same upstream state — gate, daily EV status, weekly structural reads, Category 6 confirmation, overlay states, open exposure — the panel must produce the same output, every time, with no residue of operator mood. If two operators with identical inputs would reach different tiers, the panel has leaked discretion into a layer that isn't entitled to it. The only sanctioned discretion at this layer is the documented manual throughput override, which exists precisely so that the rare legitimate exception is logged as an exception rather than laundered through the machinery as a normal decision. Everything else must replay.

The key idea

Demote the throttle to interpreter, and every layer above it becomes real.

A gate that can be overridden at deployment time is a suggestion. A structural veto that the allocator can reason around is a comment. The entire value of the upstream stack — months of evidence discipline, diagnostic machinery, capital-state doctrine — is realized only if the final layer treats its outputs as law. That is why the throttle's constitutional weakness is the system's greatest strength: it is the one design decision that makes all the other design decisions matter.