Repository navigation
Conversation
…200's bit word At 373 partner-diagonal operations of PR200's bit word, the control register is a pure copy of the passive source. These operations now read the passive data register, the copies and their copy operations go, and those pairs' partner mixes run right after the injections. Every PR200 frame, sink and alias stays. PR205's banks and PR193's assembly then give kappa = 692144604250060/10^18 (+1.088% over PR217). Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
|
Correction from the author. This PR is not a new frontier, and the served operation is not new. #210 (eumemic) already contains it. Its commit I compared only against new PR numbers and missed the head updates of #210. The package here is still an independent check of the 373-operation subset on PR200's word, but I do not claim a gain over #210. |
|
Thank you @ikeboy. This contribution is now retained on main in the scoped 100-PR review and contributor credits. Contribution: Served diagonal operations on PR200. Review and limits: Prior exact-head changed-word geometric replay and source-bound Linux full dependency job. Retains the inherited multiplication interfaces. The complete original source, notices and relevant execution evidence are retained through the pinned source archive. The review distinguishes fresh local runs, source-bound historical Linux CI and prior exact-head review; it does not claim 100 fresh full mathematical audits. Closing this historical PR as reconciled. This preserves parallel and cumulative credit without replacing the selected frontier or promoting priced research targets to completed constructions. Reviewed head: |
Change and mathematical scope
Conditional κ = 692144604250060/10^18 = 6.92144604250060e-4, under the same retained interfaces as PR200, PR205 and PR197. No hypothesis changes.
684696826673891/10^18), the best completed claim I know of.683061299399923/10^18).685800818606951/10^18, above Conditional κ = 6.84696826673891e-4: paid XOR butterflies with cascade-optimized frames #217.The bit branch binds. PR193's complex supplier (
700918443859411/10^18) has slack.The change: served diagonal operations. PR200's bit word mixes each partner pair on its data registers. The carrier register gets
x_c + x_d, and the passive register keepsx_dat all times. The word also buildsx_c + x_dagain in auxiliary registers. At 373 of these operations, the control register is a pure copy ofx_d: one copy operation makes it, and this operation uses it once.This PR reads
x_dstraight from the passive data register at those 373 operations. It deletes the 373 copies and their copy operations. To keep the passive register's frame chain nested, the pair's partner mix moves to right after the injections. The passive chain gets one 3-dim stop:line → mix → F_i → full.WEvery frame stays one of PR200's witnessed frames (24,401). PR200's 34 terminal sinks, 1,760 aliases, gauges and all other operations are unchanged. PR205's completed banks (PR197's construction) and PR193's unchanged 47-constraint assembly then apply.
research/served-diagonal-bit/PROOF.mdgives the argument.This is a conditional finite construction and composition, not an unconditional theorem or a speedup claim. The all-size hypotheses (completed core, Clifford interface, uniform weighted compilation, restored rows, fixed-tape transfer) are inherited exactly as in PR200, PR205 and PR197.
Validation
Finite checks, all exact.
research/served-diagonal-bit/verify.py --full:patch.py, which derives the served word again from PR200's frozen word and must match the frozen files;bit/served_prove.py, a complete check in the style of PR200'sbit/prove.py:W = 55,283, deficit5,808);The full run takes about five minutes on Linux, and it reproduces
certificate.jsonexactly. The workflow.github/workflows/served-diagonal-bit.ymlruns it in CI. The proof argument (why the served word computes the same columns, and why the ledger is right) is inPROOF.md; the finite checks above test it on every column.Credit
The served operation is new. It builds on:
I have not tried to stack the served lever with #217's butterflies and frames. #219 reports a higher number, but it says that R1 to R4 are open, so I compare with #217.
🤖 Generated with Claude Code