# Quantum No-Lab Validation Ladder Date: 2026-06-07 Task: td-6f39c3 ## Thesis The no-lab program should validate the Xypher quantum bridge in this order: ```text proof obligations -> deterministic adversarial row sweeps -> standard experiment replay -> public/cloud quantum traces -> apparatus work calibration -> hard d_eff experiment design ``` The target claim stays disciplined: ```text quantum information becomes state only through invariant, costed resolution ``` Not: ```text collapse mechanism proven full Born theorem complete laboratory measurement problem solved ``` ## Ladder ### 1. Proof Obligations Born weighting must be derived from admission invariants, not raw branch reward. Current proof pressure: ```text same event, same accounting, regardless of lawful refinement ``` Within power rules `mu_p(i) proportional to |c_i|^p`, orthogonal coarse-graining invariance selects `p=2`. The next proof pass should move from the power-rule/projector slice to finite effect algebras and then toward a Gleason-style statement: ```text mu(E_1 + E_2) = mu(E_1) + mu(E_2) mu(E | C_1) = mu(E | C_2) E_mu[Xi] is invariant under lawful detector refinement ``` ### 2. Deterministic Adversarial Sweeps Existing fixtures validate specific named cases. The next no-lab step is generated counterexample search over many valid state/effect rows. New harness: ```bash cd thaim-core cargo run -q --release --bin quantum-adversarial-admission-sweep ``` The harness generates deterministic qubit states and valid POVM contexts: ```text coarse context: {E, I-E} refined context: {fE, (1-f)E, I-E} alternate context: {E, g(I-E), (1-g)(I-E)} ``` It sweeps nonlinear normalized rules: ```text q_gamma(E_i | C) = Tr(rho E_i)^gamma / sum_j Tr(rho E_j)^gamma ``` Expected result: ```text gamma = 1 admits all generated rows gamma != 1 rejects under refinement/context/Xi-drift residuals raw non-Born expected-Xi winners still reject before ranking ``` This is a stronger falsification surface than a hand-picked fixture because a counterexample can arise from generated valid effects, random Bloch states, split ratios, or Xi assignments. ### 3. Standard Experiment Replay Replay textbook experiments as row contracts: ```text Mach-Zehnder phase/loss/decoherence double slit with which-path erase/readback rows Stern-Gerlach sequential basis rows Bell/CHSH effect rows quantum eraser delayed readback rows ``` The question is not whether standard quantum mechanics predicts the frequencies. It does. The Xypher question is whether one admission stack accounts for every record without hidden path state, retrocausal mutation, or raw reward authority. ### 4. Public Or Cloud Quantum Traces Use published or cloud-hardware counts as `ResolutionFrequencyRow` inputs: ```text QuantumStateRow BornWeightRow ResolutionFrequencyRow DeltaSRow WorkRow XiResolutionRow ReadbackRow EffectRow IntegrationRow AdmissionReport ``` No local lab is needed for this stage. A public trace is enough to test whether the deterministic row contract can ingest real noisy frequencies without adding a hidden selection thumb. ### 5. Apparatus Work Calibration Current harnesses use dimensionless work rows such as surprisal work: ```text W_paid >= -log2(P(selected_detector)) ``` The next calibration pass should map published detector specs into comparable work rows: ```text readout energy dark-count correction detector efficiency / loss sink erasure/reset cost timing-window cost ``` The requirement is unit discipline, not perfect metrology in the first pass. ### 6. d_eff Selection The hard claim remains: ```text d_eff >= 2 integration is necessary for state-forming measurement ``` This is the part most likely to require a custom physical experiment later. Before that, no-lab work can still sharpen falsifiers: ```text d_eff=1 apparatus selects state beyond decoherence under controlled replay decoherence-only emits reusable state records without readback/effect/work state records erase without work/thaw/accounting ``` ## Current QVAL-1 Deliverable QVAL-1 starts the program with the generated adversarial sweep. It tests the core proof pressure: ```text admission first ranking second ``` If a non-Born rule wins raw expected `Xi`, that is not enough. It must preserve effect refinement, context invariance, and expected state-admission accounting. Otherwise it is bookkeeping-dependent and cannot admit state. ## QVAL-1 Result Command: ```bash cd thaim-core cargo run -q --release --bin quantum-adversarial-admission-sweep ``` Result: ```text adversarial_admission_sweep=PASS passes=true ``` Summary: | gamma | admitted rows | raw expected-Xi winner rows | |---:|---:|---:| | 0.25 | 0 / 256 | 128 | | 0.50 | 0 / 256 | 0 | | 1.00 | 256 / 256 | 0 | | 1.50 | 0 / 256 | 0 | | 2.00 | 0 / 256 | 0 | | 4.00 | 0 / 256 | 128 | Checks: ```text born_gamma1_admitted_all=PASS born_refinement_residual_zero=PASS born_context_residual_zero=PASS born_expected_xi_drift_zero=PASS nonborn_rejected_all=PASS nonborn_refinement_residual_detected=PASS nonborn_context_residual_detected=PASS raw_nonborn_xi_winners_exist=PASS raw_nonborn_xi_winners_rejected=PASS raw_nonborn_xi_winner_count=256 ``` Reading: ```text raw branch value can prefer non-Born measures but admitted state accounting still selects Born-linear gamma=1 ``` This strengthens the no-lab route because the pass was not a single named fixture. It was a generated sweep over valid qubit states, valid effects, event refinements, alternate detector contexts, split ratios, and `Xi` assignments.