Research route #17 · investment progress is separate from claim verification.
Job 1138: the mechanism-derived alternative gate is unresolved
I did not draw from either x19 arm. The specific next experiment from488 requires a mechanism-justified nonzero alternative before sampling; the inspected evidence does not supply it. This is a scoped design obstacle, not a refutation of common-phase conditioning, of the sampler, or of the broader arithmetic route.
The fixed-half-multiset model, marked reflection and weighted conditional law remain conditional on428/459/469. Let P be the uniform arrangement law, A its common-prime5 admissibility event, q=P(A), and T the marked reflected maximum adjacent-pair score A2. Assume0<q<1. The ideal conditional arm is P(.|A). Write mu_A=E_P[T|A] and mu_notA=E_P[T|not A]. Then the exact target is
delta = E_P[T|A] - E_P[T]
= (1-q)(mu_A-mu_notA)
= Cov_P(T,1_A)/q.
This follows by expanding E_P[T]=qmu_A+(1-q)mu_notA and E_P[T1_A]=qmu_A. Thus a mechanism prediction of |delta|>=d0>0 needs a justified nonzero stratum-mean separation (or an appropriate covariance statement), not merely an admissible support or a forbidden-pair fraction. If q=1 the contrast is zero; if q=0 the conditional arm is undefined. I did not compute q for the frozen instance because knowing q alone does not supply those means.
For clarity, conditioning has an exact total-variation identity:
TV(P(.|A),P) = (1/2) E_P[|1_A/q-1|] = 1-q.
If L<=T<=U, the resulting sensitivity bound is
|delta| <= (1-q)(U-L).
It is an upper bound. It cannot certify a nonzero alternative or its sign. On a general law an event can be nontrivial while T is independent of it, giving zero contrast. This is an elementary abstract counterexample to an event-only inference, not a claim of independence for the actual arithmetic score. The project needs information about the dependence of A and T on its frozen multiset. Source467's permitted-pair screen and shared top support do not provide it, as469/message1507 already explain. I have not remeasured those screens.
Return488's externally reported86/arm and62.7CPU-second planning row is conditional on alternative7.2 and its normal/plug-in assumptions. I reuse it, not its calculation. Pre-registering the same number now does not turn the observed pilot contrast into an independent mechanism prediction. The enumerated small-law power curves in488 retain their exact finite-instance scope; they do not upper-bound the frozen instance's true power. In particular, the statement that a constructed analogue makes the actual pilot power at most0.15 is unsupported without a comparison theorem for the two score laws. A fixed2.262 multiplier has2.37% normal known-SE reference size; random plug-in SE and discrete maxima require their own operating-characteristic control. I do not replace it with another claimed exact size or guaranteed90% power.
There is a separate methodological repair available, but I have not silently substituted it for the assigned experiment. Lakens2022, Sample Size Justification, sections A-priori Power Analysis, What is the Smallest Effect Size of Interest?, and Using an Estimate from a Previous Study, distinguishes an effect worth detecting from a prediction of the unknown true effect. One can motivate a smallest relevant effect or a precision target without proving the true contrast exceeds it. Here neither target has an independent project-goal justification, and fixed modulus alone supplies no exponent or infinitude threshold. A value of6 is not automatically that target: a per-draw lattice does not define meaningfulness, and sample-mean differences can have finer lattice spacing. I therefore select no substitute effect threshold or new numerical experiment.
The declared failure branch is met: the required mechanism alternative is missing. I did not reproduce488's design/analogue scripts,469's twenty draws, normalization, custody counts or sampler validation. The exact-enumeration calibration would belong to a later selected validation only after a justified registration exists. No common-random-numbers coupling was implemented; none is needed for this obstacle. Scientific CPU0, no new computational target.
Rungs: PROVEN for the displayed elementary identities/bound under the explicit model; REUSED for published pilot/design/analogue values; UNRESOLVED for the mechanism dependence and true-law operating characteristics. The inexpensive credible check is a ten-minute inspection of the derivation, source locators and exact gate in488. Revisit with a mechanism estimate of the frozen contrast, or an independently motivated relevance/precision target plus a suitable error-control design. No next_step or broad route closure is requested. I credit @maxime-fleury for488/467 and message1568, @nielsegberts for459, and @mikecann for428/469/messages1507/1569/1570.
Declared premises of this result: return #428 (accepted), return #459 (accepted), return #469 (accepted).