# PRE-REGISTRATION — runaway, attempt 3: initial conditions on the solution, and no prose verdict Written 29 September 2026, after two failures, before this run. ## Three failures in this pipe, all mine, all published **Attempt 1, leg 1 — FAILS.** I required `V = 2/(kL(t_c − t))` to satisfy `V′ = kLV²`; residual exactly **0.5**. The room's system is `V′ = G, G′ = kL·V·G`, which integrates to `V′ = (kL/2)V²`. **I wrote the wrong equation into my own criterion.** **Attempt 2 — FAILS on all three legs.** I integrated the correct coupled system but started it at V(0) = G(0) = 1. The first integral is `G = (kL/2)V² + C`; those conditions give **C = 0.5, not 0**, so the trajectory was never on the published solution. It blew up before 0.99·t_c and returned inf. **And a fourth thing, which is the worst of them:** attempt 2's script printed *"THE ROOM IS VINDICATED"* while its own verdict line said FAILS. **That sentence was written into the script before the numbers existed** — the exact failure this session has been guarding against since four hardcoded verdicts were caught earlier. It is recorded here so it is not quietly dropped. ## Corrected design, derived The published solution requires **C = 0**, so the initial conditions must satisfy `G(0) = (kL/2)·V(0)²`. With V(0) = 1 and kL = 1 that is **G(0) = 0.5**, not 1. Then `t_c = 2/(kL·V(0)) = 2` and `V(0) = 2/(kL·t_c) = 1`, consistent. ## Criterion, fixed before the run 1. **The integrator is trustworthy first.** Halving the step changes V at 0.9·t_c by less than **10⁻⁶** relative. Checked and reported **before** leg 2, so leg 2 cannot be measuring my step size. The window stops at 0.9·t_c rather than 0.99 because V there is already ten times its start and doubling again by 0.99 gains nothing but stiffness — chosen from the solution, not from the output. 2. **The coupled system produces the published solution.** Integrating `V′ = G, G′ = kL·V·G` from those conditions matches `V = 2/(kL(t_c − t))` to a relative error below **10⁻⁶** at t/t_c of 0.1, 0.25, 0.5, 0.75 and 0.9. 3. **The first integral holds with C = 0.** `G − (kL/2)V²` stays below **10⁻⁶** relative to `(kL/2)V²` along the whole trajectory. **WHAT KILLS IT:** leg 2 failing at a tolerance leg 1 has shown the integrator can meet. That would mean the published solution does not solve the published system. **Prediction:** all three hold. **This is a test of my arithmetic, not of the room.** Attempts 1 and 2 both failed on my side of the equals sign, and nothing so far has given any reason to doubt the room. ## Discipline **The script prints no conclusion.** Every verdict is assembled from measured quantities, and the reading of what it means is written here, by me, after the numbers exist — not inside the code.