Skip to content

Re-price PR2920's two-history P104 profile under reported lifetime floors (no new kappa) - #2924

Merged
Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
Sudarsh3301:p104-two-history-lifetime-floor
Oct 16, 2026
Merged

Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
Sudarsh3301:p104-two-history-lifetime-floor

Conversation

@Sudarsh3301

@Sudarsh3301 Sudarsh3301 commented Oct 16, 2026 •

Copy link
Copy Markdown
Contributor

Summary

Audit pending PR #2920's two-history P104 price against the lifetime floors reported in the same X-only monotone model.

#2920 (head b9ea9def) finds the unique cheapest profile (cheap, unit) = (0, 208), with selected moment ≈ 0.998896 and relaxed root ≈ 0.0021042. Its search enforces #2909's endpoint curve and C+U <= 482, but no lifetime floor above the merged 208. Sagar Thakrar's mixer handoff (2245 comment 18960667, fork 743640530) reports written floors of 216 (grammar G′), 219 (shear-only G) and 234 (final-frame words). Each excludes 208 lifetimes.

floor L   cheapest (C,U)   selected moment   relaxed root
  208        (0,208)        0.9988962439     0.0021041528   #2920 control
  216        (8,208)        0.9989395970     0.0020832186   applies to #2920's model
  219       (12,207)        0.9989524538     0.0020770912   shear-only grammar
  234       (28,206)        0.9990303403     0.0020407230   final-frame words
  387      (194,193)        0.9998152606     0.0017346178   unproved relay estimate
first rejecting floor: 423 lifetimes   (232,191), moment 1.0000035

The #2920 decision stands: the restricted route still passes the optimistic screen under every reported floor. But (0, 208) is infeasible under them, and the root falls about 1% under the applicable 216 floor. A future floor rejects this two-type relaxation only at 423 lifetimes or more.

Adds only research/p104-two-history-lifetime-floor/. Credit Nick Petalas/#2920/#2909/#2896, Sagar Thakrar (floors and mixer handoff), Hiroki Kumazaki/#2353/#2889, Philippe Van Velzen, Arham Tahir/#2675/#2682, ywz65/#2614, Douglas Colkitt (ceiling.py) and the retained P104 lineage. Claim.

Evidence

git fetch --no-tags https://github.com/CrocSwap/integer-mult-bounds.git pull/2920/head
python3.12 -B scripts/research_package.py --package research/p104-two-history-lifetime-floor

Base: canonical main 3bee87dc2, priority revision 5. Live overlap was checked at 11:28 UTC with gh (open/merged PRs and the #2245, #1166 threads); #2920 had no comments or reviews.

Merge Danger

Door: two-way. One new opt-in package; predecessors, frozen inputs and the selected frontier are unchanged.

Blast Radius: scoped. The 216/219/234 floors are imported as reported (author-checked, unmerged fork), not re-proved; the 219 row is conditional on shear-only mixing. Everything else is #2920's relaxation: two named histories, separate endpoints and favorable fractional idle. No construction, new floor, supplier, kappa or human review is claimed.

Prepared by Sudarsh Chaturvedi with substantial Anthropic Claude assistance.

🤖 Generated with Claude Code

@Sudarsh3301
Sudarsh3301 force-pushed the p104-two-history-lifetime-floor branch from 50012fc to 393b342 Compare October 16, 2026 11:58
…oors

PR2920's unique cheapest profile, 0 cheap + 208 unit-step helpers, has 208
lifetimes, below the 216/219/234 floors reported in the mixer handoff for
the same X-only monotone model. Adding C+U >= L and re-optimizing with the
certified ceiling.relax engine: L=216 gives (8,208), moment 0.99894, root
about 0.0020832; the route still passes under every reported floor and the
387 relay estimate. The two-type screen first rejects at 423 lifetimes.
Floors imported as reported; no construction, new floor or kappa change.

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.

2 participants