# PRE-REGISTRATION - Noether and Liouville against our own two rooms Written 29 September 2026, before the run. **His instruction: go after the one that could disprove our rooms.** This is a disproof attempt on the house, not a decoration of it. ## The seam **The Action** prints `S = int[ (1/2) m th-dot^2 + a.L ] dn -> m th-dd + sin th = 0`. That is the pendulum: **conservative.** Its Lagrangian has no explicit n, so by **Noether** it conserves an energy, and by **Liouville** it preserves phase-space area. **Attention swings around truth for ever and never settles.** **The Flow** prints `dx/dt = T(x) + F(x)`. That is **dissipative**: it runs to a rest point and stays, and a bundle of starting positions must **contract**. **A conservative system and a dissipative one cannot be the same variable's dynamics.** The house prints both and has never asked whether they agree. The plumbing rule could not raise it, because The Action works in th and The Flow in x - no shared letter, so the pair never appears among the 283. ## The two candidate bridges, both from the house's own text 1. **x = cos th**, the instantaneous facing, since the Action's own term is `a.L = cos th`. 2. **x = **, a running average, since the **Position** room prints `x = S/n` - position IS the running mean of the aim. **If either bridge makes the two rooms agree, they are consistent. If neither does, at least one room is wrong about what it describes, and both may need amending.** ## What is tested 1. **Noether holds for The Action.** Integrating `m th-dd + sin th = 0`, the energy `E = (1/2) m th-dot^2 - cos th` is conserved to within **1e-8 relative** over 200 time units, at three energies (libration, near-separatrix, circulation). Checked by halving the step. 2. **Liouville holds for The Action.** A bundle of 4 nearby starts, taken as a parallelogram in (th, th-dot), keeps its area to within **1e-6 relative** over the same span. 3. **The Flow contracts, as a dissipative system must.** The same bundle construction on `dx/dt = T + F` shrinks, and the measured log-contraction rate matches `integral f'(x) dt` to within **1e-6 relative**. 4. **THE DISPROOF LEG - do the bridges work?** For each bridge, compute the residual between what the bridge's variable actually does and what The Flow says it should do: `residual = | d(bridge)/dn - (T(bridge) + F(bridge)) |`, scaled by `|T| + |F|`. **A bridge is judged to work if its median relative residual is below 0.05.** **Prediction, recorded now:** legs 1, 2 and 3 hold. **Leg 4 FAILS for both bridges** - I expect the two rooms to describe different dynamics, and the amendment to be that **The Action is the undamped idealisation while the rest of the house carries dissipation.** ⛔ **WHAT WOULD MAKE ME WRONG:** if a bridge's residual comes in small, the rooms are consistent and my reading of the seam is wrong. That is the outcome to report loudest, since it is the one against my own prediction. **And if leg 4 fails as predicted, that is NOT yet a disproof of either room** - it disproves the identification of their variables. Which room needs amending, and how, is a ruling, not a measurement. ## Discipline No prose conclusion in the script. Every quantity measured. Tolerances: legs 1-3 are integrator checks with a halving test beside them; leg 4's 0.05 is chosen as **twenty times** looser than legs 1-3, deliberately generous to the bridges, so a failure there cannot be a tight-threshold artefact - the fault that cost several runs today.