# PRE-REGISTRATION — is The Climb Out's 2,294 right, and the approximation beside it wrong here? Written 29 September 2026, before the run. **Closing the open item raised by entry 29.** ## The open item Entry 29 reported that applying The Climb Out's formula with its own barrier and the flow's own curvatures gives **18.6 cycles** at σ = 0.3, while the room publishes **2,294** (simulation-verified at 2,397). The exponential law checked out — the published σ=0.2/σ=0.3 ratio is 7.82 against 7.41 computed — but the prefactor was off by **123.3× at σ = 0.3 and 130.2× at σ = 0.2**, not a constant, so not a unit conversion. It was left OPEN. ## The hypothesis The room prints **two** things joined by "≈": T = (2/σ²) ∫ e^(2U/σ²) ∫ e^(−2U/σ²) ≈ π e^(2ΔU/σ²) / √(U″(bottom)·|U″(ridge)|) The first is the **exact** mean-first-passage double integral (Pontryagin, Andronov and Vitt, 1933). The second is **Kramers' asymptotic**, which requires the barrier to be large against the noise: ΔU ≫ D, with D = σ²/2. Here ΔU = 0.0721 and D = σ²/2 = 0.045, so **ΔU/D = 1.60**. That is not ≫ 1. **The asymptotic form is being evaluated far outside its own condition of validity** — the same class of error I made with the Veil's leading order in entry 31, where I tested an asymptotic law where its assumption fails. **So the prediction is that the exact integral gives the published 2,294, and entry 29's 18.6 came from evaluating the approximation in a regime it does not cover.** ## Criterion, fixed before the run 1. **The exact integral reproduces the published number.** Computing `τ = (2/σ²) ∫_bottom^ridge dy e^(2U(y)/σ²) ∫_(−1)^y dz e^(−2U(z)/σ²)` with the house's own flow (k = 1, p = 0.35, C = 0.018975) gives **2,294 within 5%** at σ = 0.3, and **17,938 within 5%** at σ = 0.2. Five percent because the room's own simulation sits 4.5% away (2,397 against 2,294), so tighter would be claiming more precision than the published pair itself carries. 2. **The integral is converged.** Halving the grid changes the result by less than **0.5%** at both σ. Reported before leg 1, so leg 1 cannot be measuring my grid. 3. **The approximation is out of its domain.** ΔU/D at σ = 0.3 and σ = 0.2 is computed and reported, and the ratio of exact to asymptotic is reported at both. **No criterion** — these are the diagnosis. **WHAT KILLS IT:** the exact integral not landing near 2,294. Then the published number does not follow from the printed exact formula either, and the open item is a real defect in the room rather than a misuse of the approximation by me. **That outcome gets reported as that, and the room gets corrected.** **Prediction:** legs 1 and 2 hold and the room is right. **Recorded plainly: I expect this to show that my own entry 29 finding was the error, not the room.** If so, entry 29 must be amended on the page. ## Discipline **No prose conclusion in the script.** Attempt 2 of entry 35 printed a verdict that its own numbers contradicted; every line of this run prints a measured quantity or a comparison against a stated criterion.