Skip to content

Price the typed quotient against direct jobs for every C (no new κ) - #4674

Merged
Mister-Meeseeks merged 2 commits into
CrocSwap:mainfrom
1337isnot1337:research/mixed-ab-controller-20261022
Oct 22, 2026
Merged

Mister-Meeseeks merged 2 commits into
CrocSwap:mainfrom
1337isnot1337:research/mixed-ab-controller-20261022

Conversation

@1337isnot1337

Copy link
Copy Markdown
Contributor

The exact mixed A/B supplier from #4654 has no recursive paid-moment discount over a direct implementation of the same main operator and all 408 independent helper jobs. This source-bound comparison proves a strict surcharge of at least 8 * 7^(-C) for every 0 <= C < 1. No higher K is claimed.

The direct main implementation uses 224 rank-seven jobs and 112 rank-one jobs. Four exact Laurent matrix identities cover both types and both raw inverse-transpose orientations, retaining signs, inverse Q powers and actual monomial normalizers. The direct implementation also pays every helper's Gaussian job. The all-exponent result follows from a normalized derivative bound, not a sampled numerical search. Fixed routing, scalar work, ownership and whole-job obligations are explicit; this compares recursive moments, not identical machine costs.

For comparison, the frozen #4610 round-three supplier does have a discount until its unique local price crossing in (0.0062073960, 0.0062073961). This is a local offer threshold, not a multiplication exponent or a universal controller ceiling.

Only research/typed-primitive-price/ is added. Read README.md and PROOF.md. Source bytes and original Git blobs are retained exactly, including compressed and CRLF inputs. The verifier independently recounts both complete emitted paid profiles and every helper rank; the underlying typed native/all-width/phase theorem is a premise, not newly re-audited.

Validation on CPython 3.12.15:

  • Registered replay passes: exact Laurent identities, source/profile authentication, outward rational endpoint brackets and charge controls.
  • Committed-byte and source-history preflight passes for all 13 package/input files.
  • git diff --check passes; scientific baseline is canonical main b58ad9ef4d4e861b0eaa5dba41742bdc072dca51.
  • No selected cold replay, new native/phase audit, formal proof assistant or hosted CI pass is claimed at opening.

Reproduce:

python3.12 -B scripts/research_package.py --package research/typed-primitive-price

User-assigned higher-K continuation changed the card scope to a prerequisite local price check before larger mixed-controller searches. The fresh overlap snapshot reports 441 open PRs and 3,598 Discussion records with complete pagination; sparse local note coverage is incomplete. This differs from #4655's geometric port budget, #4660's A-only consumer obstruction, and #4659's retained path ceiling. The B-autonomy identity is credited to the prior work rather than claimed as new.

Credit Dugongue/#4654, DaysSky, Cao Yifan/#4610, #4660, Cinna Davis and the PR4470/native/controller lineage, zxv, Jacob Alan Sussman and all retained source authors. Prepared for Cinna Davis with substantial OpenAI Codex assistance; no subagents. Apache-2.0.

No selected result or shared verifier changes. The next useful construction needs a changed native/centre recipe that restores an actual local discount, or a controller mechanism with a separately justified benefit. This result does not prove all mixed consumers impossible.

@Mister-Meeseeks
Mister-Meeseeks merged commit b21331d into CrocSwap:main Oct 22, 2026
2 checks passed
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