// ============================================================================ // NATAL-CAP 35 — THE MULTIPLICITY SPECTRUM OF THE ANCHORED OVERLAP CREDIT // (2026-08-15 — brief W7-3, successor to natal-cap-31-calm-vs-kill) // ============================================================================ // WHY. cap-31 localised the whole anchored survivor fluctuation in the overlap // credit X = Σ_r (m_r − 1)⁺ (corr(X,S) = 0.99, corr(VR,S) ≈ 0), proved the // X-limitation theorem (strikes alone cannot annihilate at any loudness), // which is proven PER LEVEL at the four levels where the ensemble maximum of // VR is enumerated: @11, @13, @17 and @19. (@19 corrected 2026-08-18: cap-19 // PART A enumerates all 9,699,690 rotations and returns V̄ = 8944.60 with // max VR = 2.293 at t = W/2, rank-from-top 0; against K = 435 and the // S̄ = 49,238.76 this file computes below, √(K·V̄·VRmax) = 2987 < S̄, so the // per-level inequality holds. natal-cap-38-loudness-driver.js states it as // the threshold S̄²/(K·V̄) = 623.1 vs max VR 2.293, margin ×271.7. The earlier // "and nowhere else" here was a false absence claim.) The theorem is still // proven at no level above @19: "from @13 upward" remains an extrapolation // off measured trends, not the theorem. It // left Assumption A restated as: the anchored overlap-credit // deficit stays below (1−ε)·S̄. It also left a warning: the anchored deficit // is invisible to pair statistics (anchored P2 reads z = −0.30 while X reads // z = −2.71), so it must live in multiplicity structure m ≥ 3. This file opens // the multiplicity m itself: the spectrum n_k = #{r : m_r = k}, anchored // against ensemble against the CRT/independent model, exactly, at four levels. // // OBJECTS. Level x, tile W = x#, natal comb N (|N| = N̄), scour primes // x < q ≤ √W (K of them), rotation t ∈ [0,W); q strikes r iff r ≡ t or t−2 // (mod q). m_r(t) = #{q : q strikes r at t}; n_k(t) = #{r ∈ N : m_r(t) = k}. // S = n_0, M = Σ_k k·n_k, X = Σ_k (k−1)⁺ n_k, P2 = Σ_k C(k,2) n_k. // Three laws in this file, all exact, all verified against brute force: // // (I1) THE u-FORM. Put μ(v) = #{q : q | v or q | v+2} for v ∈ Z. Then // m_r(t) = μ(r − t), so the ANCHOR reads μ on the natal set itself and // the ENSEMBLE reads μ on the whole integer line with a triangular // weight: E[n_k] = (1/W)·Σ_{|v| y (two primes > y would exceed W). Hence // m_r(0) = ω_s(r) + ω_s(r+2) with the hard support cap // m_r(0) ≤ 2·log W / log(x+1) — a cap the CRT model does not have. // // PREDICTIONS ON RECORD (formed on the @11/@13 pilot, before the @17/@19 runs // and before any coincidence census was computed): // P1 [proven before measuring] For any set Q of scour primes with ∏Q ≥ W+2, // the two ALIGNED patterns (all q | r, or all q | r+2) contribute exactly // zero to the anchored census, because the only multiple of ∏Q in [0,W) // is 0 and 0 is not natal. Every pair has ∏Q < W (q₁ < q₂ ≤ √W), so this // exclusion is a pure m ≥ 3 phenomenon. Predicted consequence: the // anchored aligned census at j = 3 runs far below the CRT value, and the // shortfall over super-W triples is exactly the aligned share. // P2 The anchored count of slots with ω_s(r) = i is p_nat·Π_i to a few // percent for i ≥ 2, where Π_i = #{n < W : n is a product of scour primes // with exactly i distinct ones} and p_nat = (1/4)·∏_{7≤p≤x}(p−2)/(p−1) // is the natal probability of a scour product (roughness comes free). // P3 The coincidence enhancement is a function of the two cofactor RANGES // W/d₁, W/d₂ and not of the aligned/mixed distinction: short cofactor // range ⇒ enhancement (a small multiplier is automatically rough), // long range ⇒ CRT. So the brief's "aligned coincidences are dense at // the origin" should hold for sub-W products and be swamped by P1. // P4 The m = 2 cell carries the sign change of X(0) − X̄ across the crossing, // and the m ≥ 3 cells carry large individual gaps that mostly cancel. // P5 The anchored X-deficit does NOT stay a fixed fraction of X̄: since // X̄/S̄ grows like the overlap capacity, X(0)/X̄ must flatten back toward 1 // even while β keeps descending. Predicted: the "descending through 1" // reading of X(0)/X̄ = 1.45, 0.991, 0.948 breaks at @19. // ============================================================================ 'use strict'; const T00 = Date.now(); const f = (v, d = 3) => Number.isFinite(v) ? v.toFixed(d) : String(v); function primesUpTo(n){const s=new Uint8Array(n+1),o=[];for(let i=2;i<=n;i++){if(!s[i]){o.push(i);for(let j=i*i;j<=n;j+=i)s[j]=1;}}return o;} function inv(a,m){let r0=((a%m)+m)%m,r1=m,s0=1,s1=0; while(r1!==0){const q=Math.floor(r0/r1);[r0,r1]=[r1,r0-q*r1];[s0,s1]=[s1,s0-q*s1];} return ((s0%m)+m)%m;} function msd(a){const n=a.length;let s=0,s2=0;for(let i=0;ip>=7); const W=30*basePs.reduce((a,b)=>a*b,1); const A=new Uint8Array(W); for(let r=11;rq>x&&q*q<=W); const H=qs.reduce((a,q)=>a+2/q,0); return {x,W,A,rho,N:rho.length,qs,K:qs.length,H,basePs,y:qs[qs.length-1]}; } // CRT / independent-strike model: Poisson-binomial with p_q = 2/q function crtSpectrum(L){ let poly=[1]; for(const q of L.qs){const p=2/q,np=new Array(poly.length+1).fill(0); for(let i=0;iv*L.N); } // ============================================================================ // PART 0 — CHAIN OF CUSTODY: reproduce cap-31 by its own method (full sweep of // the rotation ensemble), and carry the per-cell spectrum moments out of it. // ============================================================================ const CAP31={11:{vr:0.344,S:45,X:28,Xbar:19.26,P2:30,zX:2.25,zP2:0.45,cXS:0.481,ratio:1.4541}, 13:{vr:0.623,S:307,X:452,Xbar:456.18,P2:601,zX:-0.16,zP2:-0.16,cXS:0.908,ratio:0.9908}, 17:{vr:0.553,S:3099,X:10381,Xbar:10947.70,P2:15770,zX:-2.71,zP2:-0.30,cXS:0.987,ratio:0.9482}}; function sweep(L,KMAX){ const {W,rho,N,qs,K}=L; const Pidx=[],Pst=[],Pcd1=[],Pcd2=[];let Vbar=0; for(const q of qs){ const n=new Int32Array(q);for(const r of rho)n[r%q]++; const st=new Int32Array(q+1);for(let a=0;a=2?a-2:a+q-2; vr+=Pcd2[j][a];dd+=Pcd1[j][a]; for(let k=st[a],e=st[a+1];kMath.abs(a-b)<=tol?'PASS':'FAIL'; console.log(`CUSTODY @${L.x}: VR(0)=${f(sw.VR[0])} [${c.vr} ${ok(sw.VR[0],c.vr,0.002)}] S(0)=${sw.S[0]} [${c.S} ${ok(sw.S[0],c.S,0)}] X(0)=${sw.Xa[0]} [${c.X} ${ok(sw.Xa[0],c.X,0)}] X̄=${f(sw.mX,2)} [${c.Xbar} ${ok(sw.mX,c.Xbar,0.02)}] X(0)/X̄=${f(sw.Xa[0]/sw.mX,4)} [${c.ratio} ${ok(sw.Xa[0]/sw.mX,c.ratio,0.0005)}]`); console.log(` P2(0)=${sw.P2[0]} [${c.P2} ${ok(sw.P2[0],c.P2,0)}] z(X)=${f(zX,2)} [${c.zX} ${ok(zX,c.zX,0.01)}] z(P2)=${f(zP2,2)} [${c.zP2} ${ok(zP2,c.zP2,0.01)}] corr(X,S)=${f(cXS)} [${c.cXS} ${ok(cXS,c.cXS,0.001)}] corr(VR,S)=${f(corr(sw.VR,sw.S))}`); return sw; } // ============================================================================ // PART 1 — THE u-FORM (I1): one arithmetic function, two readings. // a(v) = #{q : q | v}, b(v) = #{q : q | v+2}, μ = a + b. // Anchored spectrum: μ read on N. Ensemble spectrum: μ read on (−W, W) with // the triangular weight w(v). Verified against the PART 0 sweep. // ============================================================================ function uform(L){ const {W,A,rho,N,qs,K}=L; const a=new Uint16Array(W),b=new Uint16Array(W); for(const q of qs){for(let j=0;j0?C[v-1]:0), wm=(W-3-v)>=0?C[W-3-v]:0, w=wp+wm; tot+=w;ens[a[v]+b[v]]+=w; ensA[a[v]]+=wp;ensA[b[v]]+=wm; // the two orientations of the u-line } tot+=N;ens[0]+=N;ensA[0]+=N; // the self-mirrored point v = −1 for(let k=0;k<=K+1;k++)ens[k]/=W; for(let k=0;k<=K+1;k++)ensA[k]/=W; for(const r of rho){anc[a[r]+b[r]]++;ancA[a[r]]+=0.5;ancA[b[r]]+=0.5;} return {a,b,C,ens,anc,ensA,ancA,totOK:tot===N*W}; } const Xof=(h)=>{let s=0;for(let k=2;k{let s=0;for(let k=j;k1e-9)bad++;if(sw.anch[k]!==U.anc[k])bad++;} chk=` [u-form vs sweep: ${bad===0?'PASS':'FAIL('+bad+')'}]`;} console.log(`\n===== @${L.x}: W=${W} N̄=${N} K=${K} (${L.qs[0]}..${L.y}) H=${f(H,4)} weight identity ΣW=N̄W ${U.totOK?'PASS':'FAIL'}${chk}`); console.log(` S(0)=${Sa} S̄=${f(Se,2)} S_CRT=${f(Sc,2)} | X(0)=${Xa} X̄=${f(Xe,2)} X_CRT=${f(Xc,2)} | X(0)/X̄=${f(Xa/Xe,4)} S(0)/S̄=${f(Sa/Se,4)} β=S(0)/S_CRT=${f(Sa/Sc,4)}`); console.log(` identity X = Σ_{j≥2}(−1)^j B_j : anchored ${f(Xa,0)} vs ${f([2,3,4,5,6,7,8,9,10].reduce((s,j)=>s+(j%2?-1:1)*Bof(U.anc,j),0),3)} ${Math.abs(Xa-[2,3,4,5,6,7,8,9,10].reduce((s,j)=>s+(j%2?-1:1)*Bof(U.anc,j),0))<1e-6?'PASS':'FAIL'}`); console.log(' k | anchored | ensemble | CRT | z(anch) | (k−1)·[A−E] | (k−1)·[A−C] | (k−1)·[E−C]'); let c2=0,c3=0,c2e=0,c3e=0; for(let k=0;k<=K+1;k++){ if(U.anc[k]===0&&U.ens[k]<1e-9&&(crt[k]||0)<1e-9)continue; const kk=Math.max(0,k-1),cA=crt[k]||0; const z=(sw&&k0)?(U.anc[k]-sw.mean[k])/sw.sd[k]:NaN; if(k===2){c2=kk*U.anc[k];c2e=kk*U.ens[k];} if(k>=3){c3+=kk*U.anc[k];c3e+=kk*U.ens[k];} console.log(` ${String(k).padStart(2)} | ${f(U.anc[k],0).padStart(9)} | ${f(U.ens[k],2).padStart(9)} | ${f(cA,2).padStart(9)} | ${f(z,2).padStart(9)} | ${f(kk*(U.anc[k]-U.ens[k]),2).padStart(11)} | ${f(kk*(U.anc[k]-cA),2).padStart(11)} | ${f(kk*(U.ens[k]-cA),2).padStart(11)}`); } console.log(` CHANNELS X-gap anchored−ensemble = ${f(Xa-Xe,2)} = m=2 cell ${f(c2-c2e,2)} + m≥3 cells ${f(c3-c3e,2)}`); console.log(` DEFICIT NORMALISED FOR THE X-FLOOR: (X̄−X(0))/S̄ = ${f((Xe-Xa)/Se,4)} [Assumption A: stays below 1−ε]`); return {crt,Xa,Xe,Xc,Sa,Se,Sc}; } // ============================================================================ // PART 3 — WHERE THE CRT MODEL BREAKS: the one-sided product truncation. // ω_s(v) = a(v) is a sum of K indicators with P(q | v) = 1/q, so the CRT model // is Poisson-binomial. The truth differs, and the difference is exactly one // thing: a set Q of scour primes can divide v < W only if ∏Q < W. Pairs always // fit (q₁ < q₂ ≤ √W ⇒ q₁q₂ < W); triples generically do not. So the CRT model // is right at j ≤ 2 and wrong from j = 3 on, by a computable missing mass. // ============================================================================ function part3(L,U){ const {W,qs,K,N}=L; const one=new Float64Array(K+2);for(let v=1;v0&&cnt<=JM){sub[cnt]+=1/prod;subCnt[cnt]++;subFloor[cnt]+=Math.floor((W-1)/prod);} for(let i=start;iW-1)break;dfs(i+1,p,cnt+1);} })(0,1,0); let ej=[1];for(const q of qs){const p=1/q,np=new Array(ej.length+1).fill(0); for(let i=0;iW} | #subsets ∏QArray.from(h.slice(0,5)).map(v=>f(v/n,4)).join(' '); console.log(` ω_s law k=0..4 truth ${shape(one,W-1)} | CRT ${pb.slice(0,5).map(v=>f(v,4)).join(' ')}`); // does the two-sided spectrum factor as the convolution of two one-sided ones? const p1=Array.from(U.ensA).map(v=>v/N),conv=new Float64Array(K+2); for(let i=0;i<=K;i++)for(let j2=0;j2+i<=K+1;j2++)conv[i+j2]+=p1[i]*p1[j2]*N; let worst=0,ws=0;for(let k=0;k<=6;k++){const d=Math.abs(conv[k]-U.ens[k]);if(d>worst){worst=d;ws=k;}} console.log(` CONVOLUTION TEST (are the two sides independent?): max|conv−ens| = ${f(worst,1)} at k=${ws}; conv n_0=${f(conv[0],1)} vs exact S̄=${f(U.ens[0],1)} (${f(100*(conv[0]/U.ens[0]-1),2)}%)`); return {one,sub,ej,subCnt}; } // ============================================================================ // PART 4 — THE ANCHORED SIDE: cofactor trichotomy (I3) and the closed form. // Every natal r is A·c, A the scour-smooth part, c ∈ {1} ∪ {primes > y}. // Closed form tested: #{r ∈ N : ω_s(r) = i} ≈ p_nat·[Π_i + Σ_{A∈S_i}(π(W/A)−π(y))], // p_nat = (1/4)∏_{7≤p≤x}(p−2)/(p−1) = λ(x)·N̄/W, λ(x) = ∏_{p≤x}(1−1/p)^{−1}. // ============================================================================ function part4(L,U){ const {W,rho,N,qs,K,y,basePs}=L; const om=new Uint16Array(W),sp=new Float64Array(W);sp.fill(1); for(const q of qs){for(let j=q;ja*(p-2)/(p-1),1); const lam=[2,3,5].concat(basePs).reduce((a,p)=>a*p/(p-1),1); const IM=6; const cof1=new Float64Array(IM+1),cofP=new Float64Array(IM+1),bad=new Float64Array(IM+1),tot=new Float64Array(IM+1); for(const r of rho){const i=Math.min(om[r],IM),c=r/sp[r];tot[i]++; if(c===1)cof1[i]++;else if(isP[c]&&c>y)cofP[i]++;else bad[i]++;} const Pi=new Float64Array(IM+1),primeRoom=new Float64Array(IM+1); for(let n=2;ny) = ${f(bad.reduce((a,b)=>a+b,0),0)} [I3 ${bad.reduce((a,b)=>a+b,0)===0?'PASS':'FAIL'}] p_nat=${f(p_nat,5)} λ=${f(lam,4)} λN̄/W=${f(lam*N/W,5)}`); console.log(' i | #{ω_s(r)=i} | c=1 | c=prime>y | Π_i | p_nat·Π_i | prime room | p_nat·(Π_i+room) | ratio'); for(let i=0;i<=4;i++){ if(tot[i]===0&&Pi[i]===0)continue; const pred=p_nat*(Pi[i]+primeRoom[i]); const tail=i===0?' n/a — the m=0 cell: r itself prime, no scour part':`${f(pred,1).padStart(16)} | ${f(tot[i]/pred,3)}`; console.log(` ${i} | ${f(tot[i],0).padStart(11)} | ${f(cof1[i],0).padStart(5)} | ${f(cofP[i],0).padStart(9)} | ${f(Pi[i],0).padStart(7)} | ${f(p_nat*Pi[i],1).padStart(9)} | ${f(primeRoom[i],0).padStart(10)} | ${tail}`); } return {om,sp,Pi,p_nat}; } // ============================================================================ // PART 5 — THE COINCIDENCE CENSUS: the brief's candidate mechanism, tested. // c(Q,ε) = #{r ∈ N : r ≡ −ε_q (mod q) ∀q ∈ Q}, ε ∈ {0,2}^Q. ALIGNED = the two // patterns all-0 (every q | r) and all-2 (every q | r+2); the rest are MIXED. // CRT value N̄/∏Q. d₁ = ∏{q : ε_q = 0}, d₂ = ∏{q : ε_q = 2}: r = d₁s, r+2 = d₂t, // so the cofactor ranges are W/d₁ and W/d₂ and the natal conditions fall on // s and t. P1 says aligned + ∏Q > W ⇒ c = 0 exactly. // ============================================================================ function coinc(L,Q,eps){ const {W,A}=L;let a=0,m=1; for(let i=0;i=W)return 0; } let c=0;for(let r=a;rMath.max(0,Math.min(NB-1,Math.floor(Math.log2(Math.max(1,z))/(lw/NB)))); const obs=[],exp=[];for(let i=0;iqs[i]),P=Q.reduce((p,q)=>p*q,1); for(let mask=0;mask<(1<W){supO+=c;supE+=e;if(al)supAE+=e;} } return; } for(let i=start;i W: obs=${f(supO,0)} CRT=${f(supE,1)} ratio=${f(supO/supE,3)} vs P1 prediction (aligned share removed) ${f(supE>0?(supE-supAE)/supE:NaN,3)} sub-W aligned enhancement ${f((aO)/(aE-supAE),3)}`); let ln=' enhancement obs/CRT by cofactor range, rows=bin(W/d₁) cols=bin(W/d₂):'; for(let i=0;i0?f(obs[i][j2]/exp[i][j2],2):' · ').padStart(8);} console.log(ln); } // ============================================================================ // DRIVER // ============================================================================ const ROWS=[]; for(const x of [11,13,17,19,23]){ const t0=Date.now(); const L=level(x); const sw=part0(L); const U=uform(L); const R=part2(L,U,sw); part3(L,U); part4(L,U); part5(L,U,2); if(x<=17)part5(L,U,3); ROWS.push({x,lnW:Math.log(L.W),...R}); console.log(` [level time ${(Date.now()-t0)/1000}s]`); } console.log('\n===== PART 6 — THE CROSSING, FACTORED ====='); console.log(' S(0)/S̄ = β · (S_CRT/S̄): β is the classical HL-vs-Mertens ratio (crosses 1 late),'); console.log(' S_CRT/S̄ is the m≥3 product-truncation factor of the rotation ensemble (falls monotonically).'); console.log(' x | lnW | S(0) | S̄ | S_CRT | β | S_CRT/S̄ | S(0)/S̄ | X(0)/X̄ | (X̄−X(0))/S̄ | classical (e^2γ/4)(1+2/L+6/L²)'); for(const r of ROWS){ const cl=0.7930547*(1+2/r.lnW+6/(r.lnW*r.lnW)); console.log(` ${String(r.x).padStart(2)} | ${f(r.lnW,3)} | ${f(r.Sa,0).padStart(6)} | ${f(r.Se,2).padStart(8)} | ${f(r.Sc,2).padStart(8)} | ${f(r.Sa/r.Sc,4)} | ${f(r.Sc/r.Se,4).padStart(7)} | ${f(r.Sa/r.Se,4).padStart(6)} | ${f(r.Xa/r.Xe,4).padStart(6)} | ${f((r.Xe-r.Xa)/r.Se,4).padStart(11)} | ${f(cl,4)}`); } console.log(`[total ${(Date.now()-T00)/1000}s]`); // ============================================================================ // OUTPUT — EMBEDDED, do not hand-edit. Regenerate: // node research/qc/embed.js --streams both research/natal-cap-35-x-multiplicity.js // invocation: node research/natal-cap-35-x-multiplicity.js // code-sha256: 23c7102472c66f2c5f07286211536d14316976f74ac88b35ea207cd109e72797 // out-sha256: c14ab734f8297682707d08878bd8388ce422c5db90994fa637359cdea9f25533 // streams: stdout+stderr // node: v22.21.0 // embedded: 2026-08-19 // elapsed: 94.8 s // ============================================================================ // CUSTODY @11: VR(0)=0.344 [0.344 PASS] S(0)=45 [45 PASS] X(0)=28 [28 PASS] X̄=19.26 [19.26 PASS] X(0)/X̄=1.4541 [1.4541 PASS] // P2(0)=30 [30 PASS] z(X)=2.25 [2.25 PASS] z(P2)=0.45 [0.45 PASS] corr(X,S)=0.481 [0.481 PASS] corr(VR,S)=0.088 // // ===== @11: W=2310 N̄=90 K=10 (13..47) H=0.7891 weight identity ΣW=N̄W PASS [u-form vs sweep: PASS] // S(0)=45 S̄=38.24 S_CRT=39.27 | X(0)=28 X̄=19.26 X_CRT=20.29 | X(0)/X̄=1.4541 S(0)/S̄=1.1768 β=S(0)/S_CRT=1.1458 // identity X = Σ_{j≥2}(−1)^j B_j : anchored 28 vs 28.000 PASS // k | anchored | ensemble | CRT | z(anch) | (k−1)·[A−E] | (k−1)·[A−C] | (k−1)·[E−C] // 0 | 45 | 38.24 | 39.27 | 1.83 | 0.00 | 0.00 | 0.00 // 1 | 19 | 35.82 | 34.28 | -2.81 | 0.00 | 0.00 | 0.00 // 2 | 24 | 13.29 | 13.11 | 3.03 | 10.71 | 10.89 | 0.19 // 3 | 2 | 2.44 | 2.89 | -0.30 | -0.89 | -1.79 | -0.90 // 4 | 0 | 0.12 | 0.41 | -0.34 | -0.37 | -1.23 | -0.85 // 5 | 0 | 0.00 | 0.04 | NaN | 0.00 | -0.16 | -0.16 // 6 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.01 | -0.01 // 7 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 8 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 9 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 10 | 0 | 0.08 | 0.00 | -0.29 | -0.70 | -0.00 | 0.70 // CHANNELS X-gap anchored−ensemble = 8.74 = m=2 cell 10.71 + m≥3 cells -1.96 // DEFICIT NORMALISED FOR THE X-FLOOR: (X̄−X(0))/S̄ = -0.2287 [Assumption A: stays below 1−ε] // ONE-SIDED ω_s over v ∈ [1,W) (the ensemble's own arithmetic, no anchoring) // j | Σ_v C(a,j) exact | Σ_{∏QW} | #subsets ∏Qy) = 0 [I3 PASS] p_nat=0.18750 λ=4.8125 λN̄/W=0.18750 // i | #{ω_s(r)=i} | c=1 | c=prime>y | Π_i | p_nat·Π_i | prime room | p_nat·(Π_i+room) | ratio // 0 | 62 | 0 | 62 | 0 | 0.0 | 0 | n/a — the m=0 cell: r itself prime, no scour part // 1 | 18 | 2 | 16 | 21 | 3.9 | 85 | 19.9 | 0.906 // 2 | 10 | 10 | 0 | 45 | 8.4 | 0 | 8.4 | 1.185 // CENSUS j=2: B_2(0) from census = 30, from spectrum ΣC(k,2)n_k = 30 [I2 PASS] CRT N̄·e_2(2/q) = 24.7 ratio 1.216 // aligned obs=19 CRT=12.3 ratio=1.540 | mixed obs=11 CRT=12.3 ratio=0.892 // ∏Q > W: obs=0 CRT=0.0 ratio=NaN vs P1 prediction (aligned share removed) NaN sub-W aligned enhancement 1.540 // enhancement obs/CRT by cofactor range, rows=bin(W/d₁) cols=bin(W/d₂): // b0 · · · · 1.88 // b1 · · · · 1.23 // b2 · · 0.62 1.43 · // b3 · · 0.57 0.95 · // b4 2.14 0.41 · · · // CENSUS j=3: B_3(0) from census = 2, from spectrum ΣC(k,3)n_k = 2 [I2 PASS] CRT N̄·e_3(2/q) = 5.0 ratio 0.402 // aligned obs=0 CRT=1.2 ratio=0.000 | mixed obs=2 CRT=3.7 ratio=0.536 // ∏Q > W: obs=2 CRT=5.0 ratio=0.402 vs P1 prediction (aligned share removed) 0.750 sub-W aligned enhancement NaN // enhancement obs/CRT by cofactor range, rows=bin(W/d₁) cols=bin(W/d₂): // b0 · · 1.63 1.74 0.00 // b1 · · 0.00 0.00 · // b2 0.00 0.00 · · · // b3 0.00 0.00 · · · // b4 0.00 · · · · // [level time 0.02s] // CUSTODY @13: VR(0)=0.623 [0.623 PASS] S(0)=307 [307 PASS] X(0)=452 [452 PASS] X̄=456.18 [456.18 PASS] X(0)/X̄=0.9908 [0.9908 PASS] // P2(0)=601 [601 PASS] z(X)=-0.16 [-0.16 PASS] z(P2)=-0.16 [-0.16 PASS] corr(X,S)=0.908 [0.908 PASS] corr(VR,S)=0.004 // // ===== @13: W=30030 N̄=990 K=34 (17..173) H=1.1468 weight identity ΣW=N̄W PASS [u-form vs sweep: PASS] // S(0)=307 S̄=310.88 S_CRT=304.28 | X(0)=452 X̄=456.18 X_CRT=449.58 | X(0)/X̄=0.9908 S(0)/S̄=0.9875 β=S(0)/S_CRT=1.0089 // identity X = Σ_{j≥2}(−1)^j B_j : anchored 452 vs 452.000 PASS // k | anchored | ensemble | CRT | z(anch) | (k−1)·[A−E] | (k−1)·[A−C] | (k−1)·[E−C] // 0 | 307 | 310.88 | 304.28 | -0.16 | 0.00 | 0.00 | 0.00 // 1 | 348 | 337.86 | 369.58 | 0.25 | 0.00 | 0.00 | 0.00 // 2 | 249 | 245.84 | 213.23 | 0.19 | 3.16 | 35.77 | 32.60 // 3 | 56 | 78.25 | 77.91 | -2.72 | -44.50 | -43.82 | 0.68 // 4 | 29 | 16.74 | 20.27 | 1.62 | 36.78 | 26.18 | -10.60 // 5 | 1 | 0.36 | 4.01 | 0.89 | 2.56 | -12.03 | -14.59 // 6 | 0 | 0.00 | 0.63 | NaN | 0.00 | -3.13 | -3.13 // 7 | 0 | 0.00 | 0.08 | NaN | 0.00 | -0.48 | -0.48 // 8 | 0 | 0.00 | 0.01 | NaN | 0.00 | -0.06 | -0.06 // 9 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.01 | -0.01 // 10 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 11 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 12 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 13 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 34 | 0 | 0.07 | 0.00 | NaN | -2.18 | -0.00 | 2.18 // CHANNELS X-gap anchored−ensemble = -4.18 = m=2 cell 3.16 + m≥3 cells -7.34 // DEFICIT NORMALISED FOR THE X-FLOOR: (X̄−X(0))/S̄ = 0.0134 [Assumption A: stays below 1−ε] // ONE-SIDED ω_s over v ∈ [1,W) (the ensemble's own arithmetic, no anchoring) // j | Σ_v C(a,j) exact | Σ_{∏QW} | #subsets ∏Qy) = 0 [I3 PASS] p_nat=0.17188 λ=5.2135 λN̄/W=0.17188 // i | #{ω_s(r)=i} | c=1 | c=prime>y | Π_i | p_nat·Π_i | prime room | p_nat·(Π_i+room) | ratio // 0 | 558 | 0 | 558 | 0 | 0.0 | 0 | n/a — the m=0 cell: r itself prime, no scour part // 1 | 310 | 7 | 303 | 73 | 12.5 | 1790 | 320.2 | 0.968 // 2 | 110 | 110 | 0 | 617 | 106.0 | 0 | 106.0 | 1.037 // 3 | 12 | 12 | 0 | 71 | 12.2 | 0 | 12.2 | 0.983 // CENSUS j=2: B_2(0) from census = 601, from spectrum ΣC(k,2)n_k = 601 [I2 PASS] CRT N̄·e_2(2/q) = 620.0 ratio 0.969 // aligned obs=283 CRT=310.0 ratio=0.913 | mixed obs=318 CRT=310.0 ratio=1.026 // ∏Q > W: obs=0 CRT=0.0 ratio=NaN vs P1 prediction (aligned share removed) NaN sub-W aligned enhancement 0.913 // enhancement obs/CRT by cofactor range, rows=bin(W/d₁) cols=bin(W/d₂): // b0 · · · · 1.51 // b1 · · · · 0.66 // b2 · · 1.49 0.93 1.27 // b3 · · 1.06 0.92 · // b4 1.44 0.64 0.76 · · // CENSUS j=3: B_3(0) from census = 182, from spectrum ΣC(k,3)n_k = 182 [I2 PASS] CRT N̄·e_3(2/q) = 214.9 ratio 0.847 // aligned obs=23 CRT=53.7 ratio=0.428 | mixed obs=159 CRT=161.2 ratio=0.986 // ∏Q > W: obs=143 CRT=186.3 ratio=0.768 vs P1 prediction (aligned share removed) 0.750 sub-W aligned enhancement 3.210 // enhancement obs/CRT by cofactor range, rows=bin(W/d₁) cols=bin(W/d₂): // b0 · · 2.99 1.47 0.45 // b1 · · 0.85 0.53 · // b2 2.53 0.30 0.31 1.91 · // b3 1.72 0.41 0.48 · · // b4 0.41 · · · · // [level time 0.313s] // CUSTODY @17: VR(0)=0.553 [0.553 PASS] S(0)=3099 [3099 PASS] X(0)=10381 [10381 PASS] X̄=10947.70 [10947.7 PASS] X(0)/X̄=0.9482 [0.9482 PASS] // P2(0)=15770 [15770 PASS] z(X)=-2.71 [-2.71 PASS] z(P2)=-0.30 [-0.3 PASS] corr(X,S)=0.987 [0.987 PASS] corr(VR,S)=0.002 // // ===== @17: W=510510 N̄=14850 K=120 (19..709) H=1.4938 weight identity ΣW=N̄W PASS [u-form vs sweep: PASS] // S(0)=3099 S̄=3614.93 S_CRT=3245.51 | X(0)=10381 X̄=10947.70 X_CRT=10578.28 | X(0)/X̄=0.9482 S(0)/S̄=0.8573 β=S(0)/S_CRT=0.9549 // identity X = Σ_{j≥2}(−1)^j B_j : anchored 10381 vs 10381.000 PASS // k | anchored | ensemble | CRT | z(anch) | (k−1)·[A−E] | (k−1)·[A−C] | (k−1)·[E−C] // 0 | 3099 | 3614.93 | 3245.51 | -2.49 | 0.00 | 0.00 | 0.00 // 1 | 5278 | 4081.48 | 5026.88 | 4.00 | 0.00 | 0.00 | 0.00 // 2 | 3820 | 4327.46 | 3797.48 | -2.68 | -507.46 | 22.52 | 529.98 // 3 | 1583 | 1958.64 | 1866.42 | -4.43 | -751.28 | -566.84 | 184.43 // 4 | 926 | 776.66 | 671.70 | 1.90 | 448.03 | 762.90 | 314.87 // 5 | 103 | 87.83 | 188.88 | 0.36 | 60.68 | -343.53 | -404.21 // 6 | 41 | 2.95 | 43.25 | 10.11 | 190.24 | -11.23 | -201.47 // 7 | 0 | 0.00 | 8.29 | NaN | 0.00 | -49.77 | -49.77 // 8 | 0 | 0.00 | 1.36 | NaN | 0.00 | -9.53 | -9.53 // 9 | 0 | 0.00 | 0.19 | NaN | 0.00 | -1.55 | -1.55 // 10 | 0 | 0.00 | 0.02 | NaN | 0.00 | -0.22 | -0.22 // 11 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.03 | -0.03 // 12 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 13 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 14 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 15 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 16 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 120 | 0 | 0.06 | 0.00 | NaN | -6.92 | -0.00 | 6.92 // CHANNELS X-gap anchored−ensemble = -566.70 = m=2 cell -507.46 + m≥3 cells -59.24 // DEFICIT NORMALISED FOR THE X-FLOOR: (X̄−X(0))/S̄ = 0.1568 [Assumption A: stays below 1−ε] // ONE-SIDED ω_s over v ∈ [1,W) (the ensemble's own arithmetic, no anchoring) // j | Σ_v C(a,j) exact | Σ_{∏QW} | #subsets ∏Qy) = 0 [I3 PASS] p_nat=0.16113 λ=5.5394 λN̄/W=0.16113 // i | #{ω_s(r)=i} | c=1 | c=prime>y | Π_i | p_nat·Π_i | prime room | p_nat·(Π_i+room) | ratio // 0 | 6813 | 0 | 6813 | 0 | 0.0 | 0 | n/a — the m=0 cell: r itself prime, no scour part // 1 | 5863 | 22 | 5841 | 257 | 41.4 | 36249 | 5882.3 | 0.997 // 2 | 1326 | 1309 | 17 | 8046 | 1296.5 | 127 | 1316.9 | 1.007 // 3 | 848 | 848 | 0 | 5273 | 849.7 | 0 | 849.7 | 0.998 // 4 | 0 | 0 | 0 | 3 | 0.5 | 0 | 0.5 | 0.000 // CENSUS j=2: B_2(0) from census = 15770, from spectrum ΣC(k,2)n_k = 15770 [I2 PASS] CRT N̄·e_2(2/q) = 16185.0 ratio 0.974 // aligned obs=7715 CRT=8092.5 ratio=0.953 | mixed obs=8055 CRT=8092.5 ratio=0.995 // ∏Q > W: obs=0 CRT=0.0 ratio=NaN vs P1 prediction (aligned share removed) NaN sub-W aligned enhancement 0.953 // enhancement obs/CRT by cofactor range, rows=bin(W/d₁) cols=bin(W/d₂): // b0 · · · · 1.11 // b1 · · · · 0.88 // b2 · · 0.99 0.99 1.03 // b3 · · 0.99 1.00 · // b4 1.09 0.88 1.03 · · // CENSUS j=3: B_3(0) from census = 7137, from spectrum ΣC(k,3)n_k = 7137 [I2 PASS] CRT N̄·e_3(2/q) = 7693.2 ratio 0.928 // aligned obs=1695 CRT=1923.3 ratio=0.881 | mixed obs=5442 CRT=5769.9 ratio=0.943 // ∏Q > W: obs=3199 CRT=4726.4 ratio=0.677 vs P1 prediction (aligned share removed) 0.750 sub-W aligned enhancement 2.285 // enhancement obs/CRT by cofactor range, rows=bin(W/d₁) cols=bin(W/d₂): // b0 · · 1.11 1.20 0.89 // b1 · · 0.87 0.85 0.59 // b2 1.08 0.82 0.94 1.04 · // b3 1.05 0.89 1.05 · · // b4 0.89 0.48 · · · // [level time 74.258s] // // ===== @19: W=9699690 N̄=252450 K=435 (23..3109) H=1.7857 weight identity ΣW=N̄W PASS // S(0)=38380 S̄=49238.76 S_CRT=41441.19 | X(0)=236625 X̄=247579.81 X_CRT=239782.23 | X(0)/X̄=0.9558 S(0)/S̄=0.7795 β=S(0)/S_CRT=0.9261 // identity X = Σ_{j≥2}(−1)^j B_j : anchored 236625 vs 236625.000 PASS // k | anchored | ensemble | CRT | z(anch) | (k−1)·[A−E] | (k−1)·[A−C] | (k−1)·[E−C] // 0 | 38380 | 49238.76 | 41441.19 | NaN | 0.00 | 0.00 | 0.00 // 1 | 83780 | 59824.54 | 75794.25 | NaN | 0.00 | 0.00 | 0.00 // 2 | 64624 | 73002.24 | 68367.12 | NaN | -8378.24 | -3743.12 | 4635.12 // 3 | 34026 | 43097.56 | 40562.03 | NaN | -18143.12 | -13072.06 | 5071.06 // 4 | 25281 | 21329.08 | 17812.15 | NaN | 11855.76 | 22406.56 | 10550.80 // 5 | 3787 | 5422.39 | 6176.91 | NaN | -6541.55 | -9559.63 | -3018.08 // 6 | 2474 | 529.22 | 1762.44 | NaN | 9723.89 | 3557.82 | -6166.07 // 7 | 98 | 6.16 | 425.67 | NaN | 551.04 | -1966.03 | -2517.07 // 8 | 0 | 0.00 | 88.86 | NaN | 0.00 | -622.01 | -622.01 // 9 | 0 | 0.00 | 16.29 | NaN | 0.00 | -130.32 | -130.32 // 10 | 0 | 0.00 | 2.66 | NaN | 0.00 | -23.90 | -23.90 // 11 | 0 | 0.00 | 0.39 | NaN | 0.00 | -3.89 | -3.89 // 12 | 0 | 0.00 | 0.05 | NaN | 0.00 | -0.57 | -0.57 // 13 | 0 | 0.00 | 0.01 | NaN | 0.00 | -0.08 | -0.08 // 14 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.01 | -0.01 // 15 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 16 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 17 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 18 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 19 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 435 | 0 | 0.05 | 0.00 | NaN | -22.59 | 0.00 | 22.59 // CHANNELS X-gap anchored−ensemble = -10954.81 = m=2 cell -8378.24 + m≥3 cells -2576.56 // DEFICIT NORMALISED FOR THE X-FLOOR: (X̄−X(0))/S̄ = 0.2225 [Assumption A: stays below 1−ε] // ONE-SIDED ω_s over v ∈ [1,W) (the ensemble's own arithmetic, no anchoring) // j | Σ_v C(a,j) exact | Σ_{∏QW} | #subsets ∏Qy) = 0 [I3 PASS] p_nat=0.15218 λ=5.8471 λN̄/W=0.15218 // i | #{ω_s(r)=i} | c=1 | c=prime>y | Π_i | p_nat·Π_i | prime room | p_nat·(Π_i+room) | ratio // 0 | 98340 | 0 | 98340 | 0 | 0.0 | 0 | n/a — the m=0 cell: r itself prime, no scour part // 1 | 109234 | 75 | 109159 | 918 | 139.7 | 717851 | 109383.0 | 0.999 // 2 | 19098 | 15778 | 3320 | 103331 | 15725.0 | 21737 | 19033.0 | 1.003 // 3 | 25245 | 25245 | 0 | 165936 | 25252.3 | 0 | 25252.3 | 1.000 // 4 | 533 | 533 | 0 | 3520 | 535.7 | 0 | 535.7 | 0.995 // CENSUS j=2: B_2(0) from census = 395426, from spectrum ΣC(k,2)n_k = 395426 [I2 PASS] CRT N̄·e_2(2/q) = 397290.2 ratio 0.995 // aligned obs=196092 CRT=198645.1 ratio=0.987 | mixed obs=199334 CRT=198645.1 ratio=1.003 // ∏Q > W: obs=0 CRT=0.0 ratio=NaN vs P1 prediction (aligned share removed) NaN sub-W aligned enhancement 0.987 // enhancement obs/CRT by cofactor range, rows=bin(W/d₁) cols=bin(W/d₂): // b0 · · · · 0.87 // b1 · · · · 1.01 // b2 · · 1.01 1.00 0.99 // b3 · · 1.01 1.00 1.01 // b4 0.87 1.01 1.00 1.00 · // [level time 0.625s] // // ===== @23: W=223092870 N̄=5301450 K=1739 (29..14929) H=2.0526 weight identity ΣW=N̄W PASS // S(0)=597475 S̄=815732.55 S_CRT=669028.80 | X(0)=6179192 X̄=6395995.55 X_CRT=6249291.79 | X(0)/X̄=0.9661 S(0)/S̄=0.7324 β=S(0)/S_CRT=0.8930 // identity X = Σ_{j≥2}(−1)^j B_j : anchored 6179192 vs 6179192.000 PASS // k | anchored | ensemble | CRT | z(anch) | (k−1)·[A−E] | (k−1)·[A−C] | (k−1)·[E−C] // 0 | 597475 | 815732.55 | 669028.80 | NaN | 0.00 | 0.00 | 0.00 // 1 | 1582579 | 1165582.82 | 1396746.38 | NaN | 0.00 | 0.00 | 0.00 // 2 | 1369538 | 1410454.72 | 1445745.12 | NaN | -40916.72 | -76207.12 | -35290.40 // 3 | 817704 | 1063505.93 | 989395.88 | NaN | -491603.86 | -343383.75 | 148220.10 // 4 | 678479 | 575779.43 | 503694.30 | NaN | 308098.70 | 524354.11 | 216255.41 // 5 | 152659 | 223074.25 | 203503.13 | NaN | -281660.99 | -203376.53 | 78284.45 // 6 | 90320 | 45135.78 | 67978.09 | NaN | 225921.08 | 111709.56 | -114211.52 // 7 | 12301 | 2159.12 | 19313.06 | NaN | 60851.26 | -42072.34 | -102923.59 // 8 | 393 | 25.32 | 4764.60 | NaN | 2573.79 | -30601.19 | -33174.98 // 9 | 2 | 0.03 | 1037.02 | NaN | 15.79 | -8280.15 | -8295.95 // 10 | 0 | 0.00 | 201.64 | NaN | 0.00 | -1814.77 | -1814.77 // 11 | 0 | 0.00 | 35.38 | NaN | 0.00 | -353.85 | -353.85 // 12 | 0 | 0.00 | 5.65 | NaN | 0.00 | -62.16 | -62.16 // 13 | 0 | 0.00 | 0.83 | NaN | 0.00 | -9.93 | -9.93 // 14 | 0 | 0.00 | 0.11 | NaN | 0.00 | -1.45 | -1.45 // 15 | 0 | 0.00 | 0.01 | NaN | 0.00 | -0.20 | -0.20 // 16 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.02 | -0.02 // 17 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 18 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 19 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 20 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 21 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 22 | 0 | 0.00 | 0.00 | NaN | 0.00 | -0.00 | -0.00 // 1739 | 0 | 0.05 | 0.00 | NaN | -82.60 | 0.00 | 82.60 // CHANNELS X-gap anchored−ensemble = -216803.55 = m=2 cell -40916.72 + m≥3 cells -175886.83 // DEFICIT NORMALISED FOR THE X-FLOOR: (X̄−X(0))/S̄ = 0.2658 [Assumption A: stays below 1−ε] // ONE-SIDED ω_s over v ∈ [1,W) (the ensemble's own arithmetic, no anchoring) // j | Σ_v C(a,j) exact | Σ_{∏QW} | #subsets ∏Qy) = 0 [I3 PASS] p_nat=0.14526 λ=6.1129 λN̄/W=0.14526 // i | #{ω_s(r)=i} | c=1 | c=prime>y | Π_i | p_nat·Π_i | prime room | p_nat·(Π_i+room) | ratio // 0 | 1784710 | 0 | 1784710 | 0 | 0.0 | 0 | n/a — the m=0 cell: r itself prime, no scour part // 1 | 2367164 | 271 | 2366893 | 3605 | 523.7 | 16300210 | 2368352.0 | 0.999 // 2 | 426060 | 231837 | 194223 | 1594672 | 231647.9 | 1335212 | 425605.7 | 1.001 // 3 | 671769 | 671769 | 0 | 4622938 | 671544.9 | 0 | 671544.9 | 1.000 // 4 | 51669 | 51669 | 0 | 356436 | 51777.2 | 0 | 51777.2 | 0.998 // CENSUS j=2: B_2(0) from census = 11044311, from spectrum ΣC(k,2)n_k = 11044311 [I2 PASS] CRT N̄·e_2(2/q) = 11078586.2 ratio 0.997 // aligned obs=5503489 CRT=5539293.1 ratio=0.994 | mixed obs=5540822 CRT=5539293.1 ratio=1.000 // ∏Q > W: obs=0 CRT=0.0 ratio=NaN vs P1 prediction (aligned share removed) NaN sub-W aligned enhancement 0.994 // enhancement obs/CRT by cofactor range, rows=bin(W/d₁) cols=bin(W/d₂): // b0 · · · · 0.78 // b1 · · · · 1.04 // b2 · · 1.00 1.00 1.00 // b3 · · 1.00 1.00 1.00 // b4 0.78 1.04 1.00 1.00 1.00 // [level time 19.363s] // // ===== PART 6 — THE CROSSING, FACTORED ===== // S(0)/S̄ = β · (S_CRT/S̄): β is the classical HL-vs-Mertens ratio (crosses 1 late), // S_CRT/S̄ is the m≥3 product-truncation factor of the rotation ensemble (falls monotonically). // x | lnW | S(0) | S̄ | S_CRT | β | S_CRT/S̄ | S(0)/S̄ | X(0)/X̄ | (X̄−X(0))/S̄ | classical (e^2γ/4)(1+2/L+6/L²) // 11 | 7.745 | 45 | 38.24 | 39.27 | 1.1458 | 1.0271 | 1.1768 | 1.4541 | -0.2287 | 1.0772 // 13 | 10.310 | 307 | 310.88 | 304.28 | 1.0089 | 0.9788 | 0.9875 | 0.9908 | 0.0134 | 0.9917 // 17 | 13.143 | 3099 | 3614.93 | 3245.51 | 0.9549 | 0.8978 | 0.8573 | 0.9482 | 0.1568 | 0.9413 // 19 | 16.088 | 38380 | 49238.76 | 41441.19 | 0.9261 | 0.8416 | 0.7795 | 0.9558 | 0.2225 | 0.9100 // 23 | 19.223 | 597475 | 815732.55 | 669028.80 | 0.8930 | 0.8202 | 0.7324 | 0.9661 | 0.2658 | 0.8884 // [total 94.579s] // ============================================================================ // READINGS (2026-08-18) — honestly calibrated // ============================================================================ // 0. WHAT THIS BLOCK CLOSES. The header put FIVE predictions on record and // then never graded them: until today this file carried no output and no // readings, and its results lived only in TODO.md item X. They are graded // below against the run pasted above. Two land, one lands in direction // only, one is unsupported, and one is REFUTED as stated. // // 1. P5 CONFIRMED, and it was the sharpest call. X(0)/X̄ = 1.4541, 0.9908, // 0.9482, 0.9558: the "descending through 1" reading breaks at @19 exactly // as predicted, and breaks upward. The replacement statistic behaves: the // deficit normalised by S̄, (X̄−X(0))/S̄ = −0.2287, +0.0134, +0.1568, // +0.2225, rises monotonically at all four levels. Four points is a lead, // not a law, and nothing here forbids it turning over at @23. // // 2. P4 REFUTED IN ITS FIRST HALF, CONFIRMED IN ITS SECOND. P4 said the m = 2 // cell carries the sign change of X(0) − X̄ across the crossing. It does // not. The total X-gap flips at @11 → @13 (+8.74 → −4.18) while the m = 2 // cell stays POSITIVE across that flip (+10.71 → +3.16); the flip is // carried entirely by the m ≥ 3 cells (−1.96 → −7.34). The m = 2 cell only // turns negative one level later, at @17 (−507.46). So the crossing is an // m ≥ 3 event, which is the OPPOSITE of what P4 predicted and, awkwardly, // the same direction as the header's own warning that the anchored deficit // lives at m ≥ 3. P4's second half holds: at @17 the m ≥ 3 cells carry // −751.28, +448.03, +60.68, +190.24 individually and sum to −59.24, i.e. // large gaps that mostly cancel. // // 3. P1 CONFIRMED IN DIRECTION, WRONG ON SIZE, AND DECAYING WITH LEVEL. The // aligned j = 3 census does run below CRT while the mixed census sits near // it: aligned/mixed ratios are 0.000/n-a @11, 0.428/0.986 @13, 0.881/0.943 // @17. But "far below" is only true at @11 and @13; by @17 the aligned // deficit is 12% against the mixed 6%, a separation of 6 points, not a // regime. The quantitative half — the super-W shortfall equals the aligned // share, i.e. 0.750 — reads 0.768 at @13 (2.4% out) and 0.677 at @17 // (10% out, and in the other direction from the trend). P1 is a good // mechanism and not yet a formula. // // 4. P2 FAILS AT @11 AND THEN TIGHTENS MONOTONICALLY. The claim was p_nat·Π_i // "to a few percent for i ≥ 2". Measured ratios at i ≥ 2: 1.185 @11 (18.5% // out — not a few percent), 1.037/0.983 @13, 1.007/0.998 @17, 1.003/1.000/ // 0.995 @19. So P2 is false where it was formed and correct to 0.5% two // levels up. The one ugly cell, @17 i = 4 with ratio 0.000, is 0 observed // against an expectation of 0.5 and carries no weight. // // 5. P3 IS UNSUPPORTED BY THIS RUN. The enhancement matrices do not organise // by cofactor range in either direction: at @13 the j = 2 row b4 reads // 1.44, 0.64, 0.76 and at @13 j = 3 the row b2 reads 2.53, 0.30, 0.31, // 1.91, neither monotone nor short-range-favouring. What the matrices DO // show is a level effect the prediction did not name: the whole j = 2 // spread collapses toward 1 as x rises — 0.41..2.14 @11, 0.64..1.51 @13, // 0.88..1.11 @17, 0.87..1.01 @19. By @19 the pair channel is CRT to within // a few percent everywhere, and the departure from CRT has moved entirely // into j ≥ 3. That is a finding, but it is not P3. // // 6. THE THREE IDENTITIES PASS EVERYWHERE THEY ARE TESTED. I2 reproduces the // anchored X exactly at all four levels (28, 452, 10381, 236625, each // against its alternating sum). I3's cofactor trichotomy records ZERO // violations at all four. The ΣW = N̄W weight identity passes at all four. // I1's u-form is checked against the brute-force sweep at @11, @13 and // @17 only ("[u-form vs sweep: PASS]"); the @19 line carries no such tag // because the sweep is precisely what the u-form replaces there. The @19 // ensemble column is therefore UNVERIFIED against brute force and rests on // the identity holding at the three levels below it. // // 7. THE @19 ROW IS A MEANS ROW, NOT A SIGNIFICANCE ROW. Every z(anch) entry // at @19 is NaN and there is no CUSTODY line for @19, because part0 runs // only where cap-31 published a reference (x ∈ {11,13,17}) and the u-form // returns ensemble MEANS, not per-rotation variances. So cap-31's headline // statistic — the anchored z(X) = −2.71 at @17 against z(P2) = −0.30, the // whole "pair statistics cannot see this" argument — is NOT extended to // @19 by this file. Anyone quoting @19 as a fourth point of that argument // is quoting a number that was never computed. // // 8. INDEPENDENCE OF THE TWO SIDES IS FALSE AT EVERY LEVEL, BY A SHRINKING // AND ONE-SIDED AMOUNT. The convolution test's n_0 against the exact S̄ // reads 2.42%, 1.63%, 1.37%, 1.01% high at @11..@19. It errs high every // time: modelling r and r+2 as independent always over-counts survivors. // // 9. WHERE THE CRT MODEL ACTUALLY BREAKS, AND IT CROSSES. The ω_s law at // k = 3 reads truth/CRT 0.0000/0.0053, 0.0028/0.0164, 0.0209/0.0321, // 0.0541/0.0481 — truth far below CRT at the small levels and ABOVE it at // @19, a crossing between @17 and @19 that no prediction anticipated. The // named mechanism is in the adjacent column: the missing mass over // super-W products, Σ_{∏Q>W}, grows 0.0069, 0.0235, 0.0398, 0.0509 at // j = 3, so the CRT model is losing exactly the subsets it has no room // for. That is P1's exclusion showing up as a global law. // // 10. COST, MEASURED, AND WHY THE NEXT LEVEL IS CHEAPER THAN IT LOOKS. Total // 80.3 s, of which @17 alone is 79.0 s and @19 is 0.96 s. The whole cost // is the two things that are switched OFF above @17: the W-rotation // custody sweep (part0, x ∈ {11,13,17}) and the j = 3 census (part5(·,3), // x ≤ 17). The u-form path that produced the entire @19 row cost under a // second. ESTIMATE, NOT A MEASUREMENT: an @23 row through the same u-form // path scales as W (23x) on the O(W log log W) legs, i.e. tens of seconds, // not hours — while an @23 CUSTODY sweep is 2.2e8 rotations x 1739 primes // and is dead. TODO.md item X says "Price the run, not the plumbing"; the // price is the sweep, and the @23 row is affordable exactly because it // cannot have one. // ============================================================================ // ============================================================================ // READINGS ADDENDUM (2026-08-19) — the @23 row // ============================================================================ // The levels array became [11,13,17,19,23]. That is the only code change, and // readings 0-9 above were written against a four-level block. Where they say // "all four levels" or quote a four-term sequence, read them as statements // about @11-@19 and take the fifth term from here. Full working, the price // derivation and the pre-registration: // research/history/staging/xchannel-at23.md // // A1. THE FIFTH POINT EXTENDS THE MONOTONE RISE, AND FASTER THAN THE FOUR // POINTS IMPLIED. (X̄−X(0))/S̄ = −0.2287, +0.0134, +0.1568, +0.2225, // +0.2658. Reading 1 said four points is a lead, not a law, and that // nothing forbade a turn at @23. No turn. But the increments are // +0.2421, +0.1434, +0.0657, +0.0433 and their ratios are 0.592, 0.458, // 0.659, so the increment ratio is NOT monotone and no saturating law // fits five points. A geometric continuation of the four-point sequence // predicted 0.246 to 0.253; the measured 0.2658 sits above that window. // The lead is stronger than the trend, which is a different fact from // the lead continuing, and neither is a law. // // A2. X(0)/X̄ RISES FOR THE SECOND CONSECUTIVE LEVEL: 1.4541, 0.9908, 0.9482, // 0.9558, 0.9661. P5 predicted exactly this flattening back toward 1 and // now has two rises rather than one. The "descending through 1" reading // is dead twice over, and the case for (X̄−X(0))/S̄ as the better // statistic is stronger than it was at four points. // // A3. THE m≥3 PRODUCT-TRUNCATION FACTOR KEEPS FALLING: S_CRT/S̄ = 1.0271, // 0.9788, 0.8978, 0.8416, 0.8202. Monotone at five points, and the // factorisation closes: β·(S_CRT/S̄) = 0.8930 × 0.8202 = 0.7324 = S(0)/S̄. // // A4. THE @23 β LEG REPRODUCES TWO EXISTING RECORDS EXACTLY. S(0) = 597,475 // and S_CRT = 669,028.80 and β = 0.8930 match paper/anchored-note.md's // §3 table row for @23 and research/OBSERVATIONS.md's exact-β table // (0.8930483 against classical 0.8884421, residual 4.606e−3; this file // prints 0.8930 and 0.8884). So the @23 row is not an unchecked new // computation: its survivor leg is a reproduction gate that passed. What // is NEW at @23 is the rotation-ensemble side, S̄ = 815,732.55 and // X̄ = 6,395,995.55, which no other instrument has computed. // // A5. THE DEFICIT HAS MIGRATED INTO m≥3 AND IS NOW DOMINANT THERE. The X-gap // against the ensemble splits −216,803.55 = m=2 −40,916.72 + m≥3 // −175,886.83, so m≥3 carries 81.1%. Across the levels where both // channels share the gap's sign the share runs 10.5% @17, 23.5% @19, // 81.1% @23. The header's warning that the anchored deficit lives at // m ≥ 3 was formed on @17 evidence where m=2 in fact carried 89.5% of // the gap; at @23 the warning is true of the gap itself for the first // time. P4's surviving half also fails here: the m≥3 cells were "large // gaps that mostly cancel" at @17 (−59.24 left of a −751.28 largest // cell, 8%), and at @23 they leave −175,804 of a −491,604 largest cell, // 36%. // // A6. NO σ_X AND NO z-SCORES AT @23, FOR THE SAME REASON AS @19. part0 runs // only where cap-31 published a reference, x ∈ {11,13,17}. Every z(anch) // in the @23 table is NaN and there is no CUSTODY line. Reading 7's // warning now covers two levels: cap-31's z(X) = −2.71 against // z(P2) = −0.30 remains an @17-only measurement, and the @23 ensemble // column is unverified against brute force exactly as @19's is. // // A7. READING 10's PRICE ESTIMATE IS GRADED AND IT LANDS. It said an @23 row // through the u-form path costs "tens of seconds, not hours". Measured: // the @23 level took 18.878 s of the run's 94.6 s, peak RSS 5.69 GB. The // estimate was formed before the array was edited and is confirmed. // Its companion claim is also confirmed: an @23 custody sweep is // 2.23e8 rotations × 1739 primes and is dead. // // A8. @23 IS THIS INSTRUMENT'S LAST LEVEL, AND THE BLOCKER IS MEMORY, NOT // TIME. The per-slot array inventory over the whole run is 25 bytes // (A 1, a and b 2 each, C 4, om 2, sp 8, isP and its sieve 2, piC 4), so // @23 needs 5.58 GB and measures 5.69 GB. At @29, W = 6,469,693,230 and // the same inventory is 162 GB. The CPU cost would be about ten minutes. // A sixth point needs a streaming or segmented rewrite, not a bigger // budget. // ============================================================================ // --------------------------------------------------------------------------- // FIGURE PROVENANCE (added 2026-08-20, readings-traceability pass). What every // figure above that the OUTPUT block does not contain verbatim actually is. // No number above was changed. // // BORROWED, verified present in the named source on 2026-08-20: // 0.8930483, 0.8884421 and the residual 4.606e-3 are research/OBSERVATIONS.md // line 515, its exact-beta table row for @23. The addendum names it. This // file prints the same two numbers to four places, 0.8930 and 0.8884, and // says so in the same sentence. // -175,804 and -491,604 are research/history/staging/xchannel-at23.md line // 217, the m>=3 cancellation figure. The increments +0.2421, +0.1434, // +0.0657 and the ratios 0.592, 0.458 are that same report's line 130. // Only the fifth increment +0.0433 and the third ratio 0.659 are new here, // and both are arithmetic on the printed (X-bar - X(0))/S-bar row. // Neither source is an embedded script OUTPUT, so the custody is a document, // not a tail. // // DERIVED IN THIS READING by arithmetic over printed values: the increment row // and its ratios (above), the geometric-continuation window 0.246 to 0.253, // the shares 81.1%, 10.5%, 23.5%, 89.5%, 8%, 36%, the 18.5% of reading on // i >= 2, and the memory estimate 5.58 GB against the measured 5.69 GB. // // ROUNDINGS of a printed value: 223092870 -> 2.2e8 and -> 2.23e8 (the @23 W, // printed in the level header line); 0.039784 -> 0.0398. // // TOKENIZER ARTIFACT, not figures: the level lists "x in {11,13,17}" and // "[11,13,17,19,23]" read as single numbers. Both are in the code above the // banner. // // DEFINITION: W = 6,469,693,230 at @29 is the primorial 29#, stated as such. // ---------------------------------------------------------------------------