Skip to content

Bound selected BIT control work uniformly across root contexts - #2310

Merged
Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
jmbarrancoml:research/selected-control-uniformity
Oct 15, 2026
Merged

Mister-Meeseeks merged 1 commit into
CrocSwap:mainfrom
jmbarrancoml:research/selected-control-uniformity

Conversation

@jmbarrancoml

Copy link
Copy Markdown
Contributor

I give an explicit control compiler and a root-independent nonchild-work bound for the pending selected BIT caller. This addresses the physical-charge handoff from PR2289/2295. Under the retained complete-stream payload and parking contracts, I derive one C_q V(n+w) bound across the allowed root contexts.

For w=ceil(e^(17/10000)), the shared Good/FallbackPair evaluator visits at most 8e^2 nodes per weight. Fixed-size matrix arithmetic and sequential access to local controls give a preparation allowance O_q(w^1800). The all-integer bound w^1800/q^(w-1) <= 2^1800*1800! absorbs this allowance into both body and high-header fibres, including width one. The compiler keeps arbitrary spectator counters outside ancestry scans. It replaces flattened-prefix address decoding with separate active-source and spectator counters.

The implementation uses an iterative stack and charged sequential reads of frozen preparation tables. After a unit reset it retains global field slots. Tests exercise Good→unit reset→Good/Pair, varying dead and live controls, complete small body fibres and stale binding/capture faults. The code tests prepared unit tables; it does not rerun the full selected matrix producers or dispatch the entire recursive payload.

The registered Python 3.12 normal and optimized replays check 82 generated coordinates, full binary request trees through depth10, 2,916 complete body records and 82,680 header fibres. The control model counts 8,775,744 bit reads and 34,737,024 head moves in its main evaluator cases; those counts are not Python timing or a full Turing-machine runtime measurement.

python3.12 scripts/research_package.py --package research/selected-control-uniformity

Review PROOF.md sections3–6 for the polynomial envelope, all-width absorption, spectator-counter restriction and parking boundary. I retain the complete physical caller and payload primitives as named dependencies. I claim no new recurrence, moment, exponent, Lean instantiation or full selected-machine verification. The next step is a joined full-ring-preparation/control/payload implementation.

Coordination: Discussion2304. I started from canonical main 0fdd7d82de03e43c9e5a1000dfbcd380d1b687d2, priority revision 4, and refreshed it before submission. The final text search covered 526 open PRs, 1,019 Discussions and 1,538 local records with complete remote pagination and no warnings. I inspected the relevant formal handoff comments; arbitrary code diffs remain outside coverage. TASK.json records remaining questions, and SOURCES.json pins the referenced interfaces.

Jose Manuel (jmbarrancoml), with substantial OpenAI Codex assistance. I retain credit for commdata2338's descriptor grammar and parking boundary, Ross Richardson's fallback algebra, the selected supplier lineage, the existing stream constructions, and Gareth Davidson's formal results and handoff. I claim no independent human verification.

@Mister-Meeseeks
Mister-Meeseeks merged commit 9c3f39e into CrocSwap:main Oct 15, 2026
50 of 61 checks passed
Mister-Meeseeks pushed a commit that referenced this pull request Oct 15, 2026
…ontrol tables

Takes the preparation/evaluator handoff in discussion #2304. PR #2310's
iterative control evaluator was checked on independently prepared unit
tables; this package joins it to PR #2234's actual full-ring selected
producer without changing either. The adapter checks the selected
request, the exact frozen-field specification and the actual low-class
child before building a table, keeps original global slots after unit
resets, and binds the fallback child's final original-row permutation
and physical atom reversal. A written lemma proves value agreement for
every admitted finite Good/Pair/unit-reset chain under the stated
producer normal-form hypotheses, and shows that numerical high fields
before the last reset need not be prepared.

Twelve selected MOVE/COPY/IDLE representatives cover every paid rank
1-7 and 42; 87 fixtures over Z/9, Z/25 and Z/27 give 801 generated-
weight comparisons and 12,642 equalities against an independent
weighted-elimination oracle. Thirteen controls reject, among others,
low-only coefficients, a discarded row permutation and an omitted atom
reversal. No mismatch in the correctly bound predecessors; no supplier
or kappa change.

Adds research/selected-control-table-binding/ only, with the consumed
PR #2234 and PR #2310 files copied unchanged under sources/ (credited to
Jose Manuel, notices retained).

Submitted by @commdata2338 with substantial AI assistance (OpenAI Codex
research and implementation; packaged with Claude Code).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X86sCSXdoDd61SDFfC1PKG
Mister-Meeseeks pushed a commit that referenced this pull request Oct 16, 2026
Takes PR #2310's released control-work handoff in discussion #2304. For
the unchanged selected BIT Good/Pair/unit-reset interpreter and root
extent e, one complete final weight vector uses at most 21e/16 value
requests, 441e/512 table reads and source-counted ring operations,
2205e/1024 popped events and 972405eL/262144 frozen-tape head moves for
page length L; each constant is sharp for the accepted grammar. The
proof couples every branch to the width loss at the same step and keeps
global field ages after resets. Under PR #2310's retained hypotheses the
sufficient preparation envelope becomes O_q(e w^3), hence O_q(w^592),
in place of its O_q(w^1800). A conditional compiler estimate only; no
child, supplier or kappa change.

Adds research/selected-control-request-mass/ only, with four PR #2310
files copied unchanged under sources/ (credited to Jose Manuel).

Submitted by @commdata2338 with substantial AI assistance (OpenAI Codex
research and implementation; packaged with Claude Code).

Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01X86sCSXdoDd61SDFfC1PKG
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