Nine layers deep: the tie-break used when two modules disagree.

The published order

When a module disagrees with another module, resolve by authority hierarchy.

The QA and maintenance rules state the sequence outright: hard plan rules, gate, throttle, cycle exposure, volatility, checklist, scorecard, structural diagnostics, then R&D. Nine positions, published in advance, covering every module in the stack. Its purpose is narrower than the four-level authority stack — that one governs conflicting signals about a decision, while this one governs conflicting outputs between workbooks. In practice the second situation is more common and less discussed, because it arrives as confusion rather than as a dilemma.

The surprising half

Expectancy evidence sits below volatility context and the checklist.

Most operators would place the Scorecard higher. It is the system's expectancy engine, its master health read, the thing that answers whether the edge is real — and it sits seventh, below the pre-entry checklist. The logic is consistent once seen: the ordering is by proximity to live capital risk, not by analytical importance. The checklist can stop a trade from happening; the Scorecard describes trades that already have. Anything that governs whether risk is deployed outranks anything that measures risk already deployed, regardless of how sophisticated the measurement is.

Diagnostics below evidence

Interpretation ranks under the thing it interprets.

Structural diagnostics sit at position two, below the Scorecard they read. This is the same principle applied one level further: the SDE and Regime Engine consume expectancy evidence and return an interpretation of trajectory, and an interpretation cannot outrank its own source. If the SDE's read and the Scorecard's numbers disagree, the numbers are what happened and the interpretation is a claim about what they mean. The diagnostic layer's value is entirely in the questions it raises for review — it has no standing to overrule the measurement it was derived from.

• The ordering is by proximity to capital risk, not by analytical sophistication.
• Anything that gates deployment outranks anything that measures deployment.
• Interpretation ranks below its source; hypothesis ranks below both.

How it gets used

The order is for confusion, not for argument.

The practical moment this list serves is mundane. The throttle says one number and a mental read of the cycle console suggests another; the volatility panel implies a coefficient that the checklist's conditions do not support; the Scorecard is green while the SDE flags deterioration. None of these is a philosophical conflict — each is an operator holding two outputs and needing to proceed. The published order means that proceeding does not require adjudication, and specifically does not require adjudication by whichever module the operator currently finds more persuasive.

The final principle

Every piece used for its intended job.

The architecture manual closes with a statement that reads as summary and functions as constraint: the checklist prevents bad entries, the capture sheet preserves evidence, CP3 organizes truth, the Scorecard measures expectancy, the Console and Throttle govern deployment, the volatility tools refine management context, SDE and Regime classify structural condition, and the EV Lab tests the future — do not blur those roles. The nine-position order is what that instruction looks like when two roles produce answers that appear to conflict. The roles were separated so that disagreement between them would be resolvable rather than paralysing.

What the order is not

Precedence is not a ranking of usefulness or trustworthiness.

The most common misreading of this list is as a quality ordering — as though R&D at position one meant the EV Lab were the least reliable module, or the checklist at position four were more sophisticated than the Scorecard at three. It is neither. The EV Lab may be the most analytically advanced instrument in the stack; the checklist is a piece of paper. The order ranks what happens when outputs conflict at the moment of a live decision, and by that measure a piece of paper that can prevent a bad entry outranks an analytical engine describing hypothetical futures. Confusing precedence with merit leads operators to argue the order upward on behalf of modules they respect, which is exactly the negotiation a published order exists to foreclose.