Skip to content
Draft
Show file tree
Hide file tree
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
140 changes: 140 additions & 0 deletions research/retained-warp-opcode-dimension/PROOF.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,140 @@
# No retained warp opcode changes dimension

This is a scoped proof about one recorded alphabet. The selected value stays
`363005817/100000000000`. The impossibility is the chart-image law on this
alphabet. Every other warp alphabet stays outside the statement.

The word is the pinned round-three centre word
`research/sol1337-contact-frontier/inputs/round3-native.json.gz`, SHA-256
`7679daa4ec6a40f56fc7f0202379a6f454310e94945aba4f48beaaaafb838d05`, the same
bytes `retained_verify.SOURCE_HASH` names. Cao Yifan / intqwq, PR #4610, is
the source of the word. The open question is the second question of PR #7102:
the 240 emitted warps preserve dimension, and a dimension-changing warp would
be a different alphabet. This note answers that question for the alphabet the
word already records.

## Hypotheses

The ambient space is `GF(2)^9`. `canonical["h"]` is 9. Every chart basis
vector is an integer in `0..511`, and the retained span test makes the basis
independent. The dimension of a chart is the length of that basis.

The recorded alphabet is `canonical["warps"]`. It is a list of 72 records
whose `op` fields are exactly `1..72` in that order. Record `k` stores a
source chart `from_frame`, a destination chart `to_frame`, and nine columns
`G_columns`. Those columns define the linear map `G_k` by

```text
G_k(x) = xor of column i over each bit i set in x.
```

The retained inverse opcode `-k` is the column-wise inverse used by
`model.inv`: the unique vectors sent to the standard basis. The emitter and
`model.graph` both keep these 144 signed maps. A free-word symbol, including
the dual label that adds `1000000` in `retained_verify.emit`, is not another
spatial map.

The chart-image law is the native monomial rule. Opcode `σ` admits an ordered
pair of charts `(A, B)` only when `G_σ` carries the span of `A` onto the span
of `B`. That is the check `model.build` states as “original native warp is
monomial”, and the check `model.build_pauli` states as “native chart image”.
`retained_verify.replay` does not apply this law: after editing the free word
it assigns the role’s chart to the event’s destination. An assignment that
fails the chart-image law is outside the alphabet. This note does not
construct one.

## The dimension lemma

Let `G` be a bijective linear map on `GF(2)^9`, and let `v_1, ..., v_d` be a
basis of a subspace `V`. If a subset of `G(v_1), ..., G(v_d)` sums to 0, apply
`G^{-1}`: the same subset of `v_1, ..., v_d` sums to 0, so the subset is
empty. The image vectors are therefore independent, and they span `G(V)`.
Hence `dim G(V) = dim V`.

Every subspace is covered, including a chart that is not a recorded endpoint
and a chart that none of the 240 emitted warps uses.

## Census of the 72 maps

The checker evaluates each of the 72 column lists on all 512 vectors. Each
list is additive, has column rank 9, and has 512 distinct images, so each
`G_k` is a linear automorphism. Each inverse agrees with `model.inv`, is a
two-sided inverse, and is an automorphism as well. The domain of this census
is the whole list `canonical["warps"]`. The opcode check rejects any proper
sublist, so a sample of the records is not this census.

Opcode 1 is the named map. Its columns are

```text
[1, 2, 4, 8, 64, 19, 104, 197, 325].
```

It joins frame 292 to frame 293. Both charts have dimension 4, and `G_1`
carries the span of frame 292 onto the span of frame 293. The two spans are
distinct: their Fourier rank, in the sense of `model.fourier_rank`, is 4.
Opcode 1 changes the chart and preserves the dimension.

The same endpoint check on all 72 records gives 50 opcodes of dimension 4 and
22 of dimension 5. Every recorded pair is a chart image, so the monomial
Fourier rank between the image and the destination is 0. No recorded opcode
has unequal endpoint dimensions.

## Unequal pairs are inadmissible

Suppose `G_σ(span A) = span B` for a retained signed opcode. The lemma gives
`dim A = dim B`. An unequal pair fails the chart-image law for every retained
opcode.

Frame 1 is the unique dimension-9 chart, and its span is the whole space.
Roles `112..223` end on 112 distinct charts of dimension 8, each a strict
subspace of frame 1. The Fourier rank of any such chart against frame 1 is 1,
the same width a transport would pay. For every one of the 144 signed maps and
every one of these 112 charts, the image has dimension 8 and Fourier rank 1
against frame 1, so the image is not frame 1. No retained opcode admits a
warp from a Y exit onto frame 1.

The 240 emitted events never use a dimension-8 chart. Their source-dimension
multiset is `{0: 72, 1: 24, 4: 50, 5: 22, 9: 72}`. Looking at some of those
events cannot see a dimension-8 pair. The prohibition on that pair is the
lemma, applied to the census of the 72 maps.

## Products

A composite of bijective linear maps is bijective and linear, so the lemma
applies to every finite product of retained opcodes and inverses. Along a walk
whose steps are admitted chart images, the dimension is constant. No such walk
carries a dimension-8 Y exit onto frame 1. The checker evaluates the composite
`G_1 ∘ G_2` as a fixture of that lemma. It does not enumerate the group
generated by the 72 maps.

## The emitted events, recomputed

This recount is an exhaustive census of a second finite list. It is the list
PR #7102 already used. It is recorded so the off-endpoint uses are visible,
and it is not the reason an unused unequal pair is forbidden.

There are 240 emitted warp events, each on a live role. Each opcode is one of
the 144 signed retained maps, each event preserves dimension, and each event
is a chart image of its opcode. Of the 240, 72 land on that opcode’s oriented
recorded endpoints. Those 72 are exactly the events of dimension 4 or 5.
The other 168 land on other charts, with source dimensions
`{0: 72, 1: 24, 9: 72}`. The alphabet is used at dimensions other than its
recorded endpoint dimensions.
`canonical["warp_edges"]` is exactly the set of those 240 events collapsed to
180 classes `(sign, opcode, source, destination)`, and each class preserves
dimension.

## What this statement leaves open

The selected frontier stays `363005817/100000000000`. Opcode 1 is a chart
change inside one alphabet, not a selected-word decoder and not a new κ.

The restored Y-exit round trip priced by PR #7102 is unchanged. Accepting the
one-way move onto frame 1 as an endpoint is a different contract. Same-
dimension relocation of a Y exit, by a retained opcode whose image happens to
be another rank-8 chart, is still open; the lemma says the image has dimension
8 and says nothing further about which chart it is.

A warp alphabet other than these 72 maps and their inverses can still contain
a dimension-changing map. The free-word replay can still be given a chart
assignment that ignores `G`. Both of those objects are outside the hypotheses.
85 changes: 85 additions & 0 deletions research/retained-warp-opcode-dimension/README.md
Original file line number Diff line number Diff line change
@@ -0,0 +1,85 @@
# No retained warp opcode changes dimension

**Scoped proof on one recorded alphabet. The selected κ stays
`363005817/100000000000`.**

On the pinned round-three centre word, the recorded warp alphabet is the 72
maps in `canonical["warps"]` together with the inverses `model.inv` retains.
Each map is a linear automorphism of `GF(2)^9`. Under the native chart-image
law, an automorphism admits only equal-dimension chart pairs. Opcode 1, the
named example, joins frame 292 to frame 293, both of dimension 4. No retained
opcode admits a warp from a dimension-8 Y exit onto the dimension-9 frame 1.
A finite product of retained opcodes preserves dimension as well.

The argument is in [PROOF.md](PROOF.md). The checker recomputes the census
and the chart images. It does not trust the counts above as inputs.

## Scope and evidence

Catalogue card `complex-rank-shape-decoder`, priority revision 8, base
commit `e54523517d0f03e58352e647c403a5af3e525a57`. Scope status:
**changed**. The card asks whether a decoder or coframe change improves the
full signed moment. This package does not build that decoder. It answers the
second open question of PR #7102, on the same #4610 round-three word. The
reason is in [TASK.json](TASK.json).

Evidence class: a proof that every bijective linear map preserves subspace
dimension, together with an exhaustive census of the 72 recorded maps. The
240 emitted warps are a second exhaustive census, recomputed from the event
list. PR #7102 already records that those 240 preserve dimension. The recount
splits them into 72 endpoint uses, at dimensions 4 and 5, and 168 other uses,
at dimensions 0, 1, and 9. The Y-exit prohibition is the lemma, because those
240 events never use a dimension-8 chart.

Source pins are in [SOURCES.json](SOURCES.json). The catalogue and selected
record are provenance at the base commit. Full Git history is needed to
resolve them. The word, the retained emitter, and the contact-graph model are
live inputs.

## Reproduce

Python 3.12, standard library, from the repository root:

```sh
python3 -B research/retained-warp-opcode-dimension/verify.py
python3 -O -B research/retained-warp-opcode-dimension/verify.py
python3 -I -W error -B research/retained-warp-opcode-dimension/verify.py
python3 -B scripts/research_package.py --package research/retained-warp-opcode-dimension --check-only
python3 -B scripts/research_package.py --package research/retained-warp-opcode-dimension
```

The marker is `PASS_RETAINED_WARP_OPCODE_DIMENSION` with `opcodes=72`,
`dimension_changing_opcodes=0`, `opcode_1=292->293 dim=4`,
`y_exit_to_frame1_admitted=0`, and `emitted_warps=240`.

The positive fixture is opcode 1: frames 292 and 293, dimension 4, with the
recorded columns and a chart-image match. Controls reject a column of opcode
1 zeroed out, a 71-opcode sublist, the pair (Y exit, frame 1) under opcode 1,
a same-dimension chart outside that image, a sample of ten emitted events as
a stand-in for the alphabet, and a width-0 reading of the one-dimension gap.
Not run: `model.build` on the full contact graph, a physical caller, Lean,
Docker, and `make frontier7000-verify`.
`scripts/research_selected_baseline.py` is a binding check only.

## Open next questions

1. Which dimension-8 charts, if any, does opcode 1 carry a Y-exit chart onto?
The dimension lemma says the image has dimension 8, so the image is not
frame 1. It can still equal another rank-8 chart. A warp onto that chart
would move the exit with no paid transport, inside this alphabet. The
first step is to map the exit basis of Y role 112 by opcode 1’s columns
and test the image against the other 111 Y-exit spans. Nothing algebraic
blocks that census; this package stops once unequal dimensions are
excluded. The one-way acceptance of frame 1 as an endpoint is a different
contract and is not this question.

## Credit

The word is Cao Yifan / intqwq, PR #4610. The emitter is the retained centre
extension. The chart-image law and `model.inv` are the contact-graph model,
which credits PR #4672, PR #4654, PR #4610, and the PR #4470 lineage. The
question is the second open question of PR #7102. The Y-exit geometry follows
the round-three indexing in `research/y-exit-rank-lift-obstruction`, which
credits discussion #1166 comment 19133861, Chris R / Clear1nsight.
Prepared by unpaidmoon with substantial Cursor Grok assistance. Apache-2.0.
No independent human review or proof-kernel check is claimed.
36 changes: 36 additions & 0 deletions research/retained-warp-opcode-dimension/SOURCES.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,36 @@
{
"schema_version": 1,
"base_commit": "e54523517d0f03e58352e647c403a5af3e525a57",
"files": [
{
"path": "certificates/selected-result.json",
"usage": "provenance",
"git_blob": "75d77b328ea02ded535e7f23aee5b1e55d0bb1a8",
"sha256": "154c3d30455d778fd3343b6b4851f0fa4665f0f7625b0adb3675e2bafbbf6bf1"
},
{
"path": "docs/research/agent-task-catalog.json",
"usage": "provenance",
"git_blob": "3404fbdd5e7529a0674eef26114b5071e4586cec",
"sha256": "ead6cdc9d0af6d58e8a49744d8ec8c7c15e13d2ba93c276c434bcf9b51b93a81"
},
{
"path": "research/sol1337-contact-frontier/inputs/round3-native.json.gz",
"usage": "input",
"git_blob": "46d929df83817b85a5387bcdb9745e0627d6366e",
"sha256": "7679daa4ec6a40f56fc7f0202379a6f454310e94945aba4f48beaaaafb838d05"
},
{
"path": "research/sol1337-contact-frontier/vendor/model.py",
"usage": "input",
"git_blob": "a69138e4014424235345f3301fbe6b097caa1b8c",
"sha256": "7cc29edccb3679b2a5c9defeeb0a2ec6bcef9b01b97c2390bebc48f366395f1f"
},
{
"path": "research/sol1337-contact-frontier/vendor/retained_verify.py",
"usage": "input",
"git_blob": "357cbff9874ce920c5b5ad839fd47d3f9c728bcb",
"sha256": "f5a67954f19dddc688460b91e5c1729d7820755c53313159ed136e4447021214"
}
]
}
75 changes: 75 additions & 0 deletions research/retained-warp-opcode-dimension/TASK.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,75 @@
{
"schema_version": 1,
"task_id": "complex-rank-shape-decoder",
"priority_revision": 8,
"base_commit": "e54523517d0f03e58352e647c403a5af3e525a57",
"original_card": {
"id": "complex-rank-shape-decoder",
"question": "Can a legal decoder or coframe change improve the full signed moment, including changed helper targets?",
"starting_points": [
"docs/research/current-priorities.md",
"docs/research/research-family-status.md",
"certificates/selected-result.json",
"docs/research/pr-intake-evidence/pr5599-completion/PROOF-REVIEW.md",
"docs/research/recursive-invoice-preflight.md"
],
"methods": [
"Construct and independently check one source-bound instance of the stated first experiment.",
"Seek an exact counterexample or target-specific complete-budget obstruction to that same proposed change."
],
"acceptance_gate": "A source-bound arbitrary-input identity and legal producer/consumer boundary with the complete paid and productive target invoice, including restoration, ownership, native alignment and extra charges. Price under the actual current model; historical ceiling.py assumptions must be established before use. A favorable optimistic relaxation keeps a route open but is not a completed supplier. A scoped local result remains useful; expansion requires the family-level budget gate.",
"first_step": {
"command": [
"python3",
"-B",
"scripts/research_selected_baseline.py"
],
"expected": "PASS_RESEARCH_BASELINE_BINDINGS_ONLY",
"inspect": "Read revision 8, the selected review and live exact-source handoffs. Then follow this card's bounded first_experiment.",
"scope": "Source-bound orientation only, not a supplier replay or theorem proof."
},
"first_experiment": "Price one complete changed itinerary with both paid and productive histograms. Check #5693/#5698 for overlap and their exact hypotheses before another coframe search. Use the signed invoice preflight where its boundary contract agrees; an inconclusive concave comparison is not a power crossing. Stop if an optimistic complete budget fails at the chosen target or native delivery remains unspecified.",
"bottleneck_link": "Judge the actual current supplier contract and complete signed target/paid bill; frozen-profile ceilings require an applicability proof.",
"allowed_variations": [
"One exact source head, helper/cut region or quantified proposition and changed hypothesis; check live handoffs first. Different seeds are not different scopes."
],
"scope_change_triggers": [
"Changing source head, helper/cut domain, operation alphabet or caller boundary without recording the new scope.",
"Treating an archived proposal or local numerical pass as complete supplier admission."
],
"overlap_terms": [
"coframe",
"signed moment",
"5693",
"5698"
],
"existing_checks": [
{
"scope": "Selected PR5599: complete helper-subspace construction and independent caller replay; five framework assumptions remain.",
"evidence": "docs/research/pr-intake-evidence/pr5599-completion/PROOF-REVIEW.md",
"remaining": "The next experiment must identify uncovered source-specific work. Successor archive intake does not transfer this admission.",
"status": "reviewed",
"source_commit": "09dd0965711ca7eeb169057a420abab1c9052137",
"selected_record_sha256": "2a0220e866417774606e0f991e41b9c1d26b1f43b97857238cf478dbc7d330b5"
}
]
},
"scope_status": "changed",
"actual_scope": "On the pinned round-three centre word, every retained warp opcode in canonical['warps'] is a linear automorphism of GF(2)^9, and the native chart-image law therefore admits only equal-dimension chart pairs. Opcode 1 joins frames 292 and 293 at dimension 4. No retained opcode, and no finite product of retained opcodes, admits a warp from a dimension-8 Y exit onto frame 1.",
"scope_change_reason": "The card asks for a decoder or coframe change that improves the signed moment. This package does not build that decoder. It answers the dimension question left open by PR #7102, on the #4610 round-three alphabet rather than on a new opcode alphabet or the selected #7000 word.",
"bottleneck_link": "A favourable complex itinerary has to clear a complete signed bill. PR #7102 charges a transport for every dimension change and leaves open whether a retained warp could move a Y exit to frame 1 for free. Under the recorded alphabet and the chart-image law, that move is inadmissible, so the route stops before a decoder or a kappa claim.",
"budget_minutes": 60,
"stop_rule": "Stop once every retained opcode is shown to preserve dimension under the chart-image law, including every Y-exit pair with frame 1. Do not construct a new opcode, do not accept the one-way frame-1 endpoint, and do not claim a kappa change.",
"outcome": "negative-result",
"evidence": "python3 -B research/retained-warp-opcode-dimension/verify.py prints PASS_RETAINED_WARP_OPCODE_DIMENSION. PROOF.md records the dimension lemma, the 72-map census, and the separate recount of the 240 emitted events. The checker requires opcode 1's columns and rejects a singular alteration, a 71-opcode sublist, and the pair (Y exit, frame 1).",
"author": "unpaidmoon",
"ai_assistance": "Substantial Cursor Grok assistance with the proof and the checker. The word, the retained emitter, and the chart-image law are predecessor source.",
"open_questions": [
{
"question": "Which dimension-8 charts, if any, does opcode 1 carry a Y-exit chart onto?",
"why_it_matters": "The dimension lemma puts the image in dimension 8, so the image is not frame 1. The image can still equal another rank-8 chart. A warp onto that chart would move the Y exit with no paid transport, inside this alphabet.",
"first_step": "Map the exit basis of Y role 112 by opcode 1's columns and test the image span for equality with each of the other 111 Y-exit spans.",
"blocked_by": "Nothing algebraic. The census is finite and was left unrun because this package stops once unequal dimensions are excluded."
}
]
}
23 changes: 23 additions & 0 deletions research/retained-warp-opcode-dimension/VERIFY.json
Original file line number Diff line number Diff line change
@@ -0,0 +1,23 @@
{
"schema_version": 1,
"status": "ready",
"python": "3.12",
"requirements": [],
"inputs": [
"certificates/selected-result.json",
"research/sol1337-contact-frontier/inputs/round3-native.json.gz",
"research/sol1337-contact-frontier/vendor/model.py",
"research/sol1337-contact-frontier/vendor/retained_verify.py"
],
"commands": [
{
"argv": [
"{python}",
"-B",
"{package}/verify.py"
],
"timeout_seconds": 120,
"checks": "Census all 72 retained warp maps and their inverses, require the chart-image law on opcode 1 and on every Y-exit pair with frame 1, and reject a singular column, a 71-opcode sublist, and a ten-event sample."
}
]
}
Loading