Skip to content

Birth-read reuse on shrunk saturated frames: conditional κ = 1.1239534209971e-4 - #127

Draft
GamingPuzzled wants to merge 8 commits into
CrocSwap:mainfrom
GamingPuzzled:research/shrunk-birth-reuse
Draft

GamingPuzzled wants to merge 8 commits into
CrocSwap:mainfrom
GamingPuzzled:research/shrunk-birth-reuse

Conversation

@GamingPuzzled

Copy link
Copy Markdown
Contributor

Draft: the package runner passes locally on Python 3.11. Fork CI is running.

Under PR124's retained hypotheses,

κ = 11239534209971/10¹⁷ = 1.1239534209971×10⁻⁴

PR κ This PR is above it by
#126 (best open at submission) 1.11192e-4 1.08%
#125 1.10592e-4 1.63%
#124 1.10383e-4 1.82%

The complex side still binds.

This composes two orthogonal improvements on #117's DAG:

More and larger deferred frames give the reuse more recipients.

h=24 complex #124 this
deferred readouts 3,020 4,706
reuses 2,106 2,108
R 26,599 26,597
W 115,865,904 115,857,808
complex saving 11040781/10¹¹ 11242073/10¹¹

Checks. Run python3 research/shrunk-birth-reuse/run_checks.py. It pins #125's DAG, profile and certificate by SHA-256. In a temporary tree it then runs:

Two harness changes from #124:

  • RLIMIT_AS is skipped on macOS, where it can't be set.
  • The frame stage's caps are raised to 120 s and 2 GiB, because the shrink and saturated placement are heavier.

Scope. This is a finite conditional witness. All of #124's and #125's inherited hypotheses remain assumptions. The branch contains #125 (and #114) because it builds on them.

Credit: James Chang/jamesyc (#124 reuse operation, lemma, matcher, bill, audit and assembly, with OpenAI assistance); eumemic (#117, #114); Rohan Arun (#118); Avi Eisenberg/ikeboy (#110); Swapnil Jain; icekylinx; Zhihao Chen and all inherited contributors. The composition and the shrunk-frame step were prepared by Joel Pulikkan with Anthropic Claude assistance.

🤖 Generated with Claude Code

icekylinx and others added 8 commits October 9, 2026 07:11
…78e-5

Under PR104's opposite-bank rule every residual is one child, so deferred
readouts pay on both networks. Bit: Swapnil Jain's round-seven deferred word,
unchanged (stopped saving 1.2402e-4). Complex: PR104's h=24 network with lifted
binary frames and deferred readouts on 8,893 of 44,918 roles (8.58628e-5).
Replay with arbitrary scratch, exact F2 frame checks, exact moments and the
47-constraint assembly give kappa = 429239/5000000000.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…0591581647e-4

On PR114's saturated deferred construction over PR117's immutable DAG,
one forward pass replaces a lifted frame by the minimal nondegenerate span
of its label and predecessor frames whenever that merges increments of the
role chains holding it. 16,031 frames shrink; roles, matching, rank mass
and 4,706 deferred readouts are unchanged. Complex saving rises from
109305097/10^12 to 110616162/10^12, giving kappa 110591581647/10^15.

All PR114 checks are retained: arbitrary-scratch replays, exact F2 chain
nesting and nondegeneracy on the shrunk frames, the independent literal
reflection audit, exact moments, 47 constraints and 7 margins. A new
control rejects PR114's unshrunk profile at the new saving.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
…534209971e-4

Apply PR124's compensated birth-read slot reuse (James Chang) to PR125's
frames: PR114 saturated placement (4,706 deferred readouts) plus the
shrunk-frame pass (16,031 frames). PR124's matcher finds 2,108 legal
reuses, leaving R = 26,597. Complex saving 11242073/10^11 gives kappa
11239534209971/10^17.

build_frames.py reconstructs the shrink and saturated placement without
upstream imports and matches PR125's complete histogram bin for bin.
PR124's lemma, readout bill, forward/reflected literal frame audit and
exact assembly run unchanged apart from the pinned profile; the next
complex grid point is now also rejected. The Linux memory cap is skipped
on macOS, where RLIMIT_AS cannot be set.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants