memory: lambda_total is 20.560 on 6909d15, and has not moved - #28
Open
austinamissah wants to merge 2 commits into
Open
memory: lambda_total is 20.560 on 6909d15, and has not moved#28austinamissah wants to merge 2 commits into
austinamissah wants to merge 2 commits into
Conversation
02-lambda.md's 23.29 was measured on 02146ca and is stale. Replaces it with a measurement on upstream/main 6909d15: lambda_total 20.560 (95% CI 18.007-23.016), n=199 non-control trials, decomposition 8.932 / 8.193 / 3.435, rho 0.569, var/mean 1.005 and 1.037. The finding is stability. Against 801dd20's 20.04 the difference is +0.525, 95% CI -2.626 to +3.632, with 37% of bootstrap resamples at or below zero. Eight accepted submissions moved lambda by nothing measurable. Clean seeds remain isolated (Welch t +0.77 / -0.09). Restructures the note so the current figure leads; the n=700 material is kept for the A-vs-B discrimination it uniquely establishes. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
The n=199 sweep in 02-lambda.md was measured on 6909d15, the 1,154-qubit trailmix/dialog circuit. That construction is now legacy, gated behind SUB4_LEGACY_POINT_ADD, with pingpong_div.rs shipping at 1,278 qubits, so the section heading 'The current figure' no longer holds. Retitles it 'The measurement' and adds a scope note. Records for orientation that the same quantity on the ping-pong circuit is about 18.75 classical mismatches per draw (n=8, sd 6.0, at SUB4_PP_ROUNDS=698 / SUB4_PP_ROUNDS_MUL=696), against 17.126 for trailmix; not directly comparable, but the same order, and neither admits a clean seed by luck. No measured figure is changed. The method gates, including the sudo env-strip gate filed as issue Layr-Labs#23, apply to any sweep on this harness.
austinamissah
force-pushed
the
memory/lambda-current
branch
from
August 23, 2026 15:47
5af248d to
911d0bc
Compare
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
02-lambda.md's λ_total of 23.29 was measured on02146caand is stale. This replaces it with a measurement onupstream/main6909d15— the head whosesrc/tree is identical to accepted submissioned4b529.λ_total = 20.560 (95% CI 18.007–23.016), n=199 non-control trials. Decomposition λ_classical_only 8.932, λ_both 8.193, λ_phase_only 3.435; ρ(c,p) = 0.569. var/mean 1.005 classical and 1.037 phase — Poisson, no overdispersion.
P(clean)moves from 7.7e-11 to 1.2e-9, and the phase channel's standalone cost from 175× to 31×.λ has not moved materially since
801dd20(20.04). The difference is +0.525, 95% CI −2.626 to +3.632, with 37% of bootstrap resamples at or below zero. Eight accepted submissions and a better score have not moved λ measurably in either direction. That stability is the finding — it means planning against a moving λ, in either direction, is planning against noise.Clean seeds remain isolated: 99 contiguous nonces and 100 at 2^40 stride are indistinguishable (Welch t = +0.77 classical, −0.09 phase). Grinding near a known-good nonce still buys nothing.
Method gates — all four passed before the sweep was trusted
6909d15build reproducesops.binmd5ef30945f3afcb369192ea32897232d2fat 0/0/0.200321420125, not801dd20's62000008397024.sudo'senv_resetstrippingSUB4_TAIL_NONCEbeforebuild_circuitsees it.801dd20figure from that sweep's raw data before being pointed at this one.Raw data and full method:
austinamissah/ecdsa-circuit-optimization—docs/lambda-6909d15.md,docs/data/lambda-sweep-6909d15.tsv.Scope — and a smaller version on offer
This restructures
02-lambda.mdso the current figure leads and the n=700 material is demoted to what it uniquely establishes (the A-vs-B discrimination). If you'd prefer a minimal edit — just the number,P(clean), and the 175× → 31× correction, leaving the structure alone — say so and I'll cut it down.One line was deliberately left unchanged: "roughly once per 1,100 inversions" in What this means. Its derivation isn't stated in the note; on the λ_both reading it survives unchanged (9024/8.193 = 1,101), but if it came from elsewhere it may now be stale.
Need help on this PR? Tag
@codesmith-botwith what you need. Autofix is disabled.