Skip to content
Merged
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
2 changes: 1 addition & 1 deletion README.md
Original file line number Diff line number Diff line change
Expand Up @@ -294,7 +294,7 @@ soundness. See `spec/VERSIONING.md`.

Seeded from an earlier proposal draft; this copy is the living document. The
formal-design packages FD0–FD7 are closed, and their artifacts are superseded by
`spec/`. The proposal, an archived internal note, and an in-preparation
`spec/`. The proposal, an internal note, and an in-preparation
manuscript are not shipped here; the calculus is stated so that its normative
content does not depend on them (`REFERENCES.md`).

Expand Down
6 changes: 3 additions & 3 deletions REFERENCES.md
Original file line number Diff line number Diff line change
Expand Up @@ -24,7 +24,7 @@ what each shorthand denotes and where (if anywhere) it can be obtained.
repository's own evidence.
- **manuscript** — synonym for *code-deformation* where the citation names a
theorem number as it appears in the manuscript (e.g. "manuscript Theorem 7").
- **an archived internal note** — an internal working note that has been
- **an internal note** — an internal working note that has been
superseded by `spec/` and `SCOPE.md`. Citations are retained as historical
provenance only and are explicitly marked *not a normative dependency*.
The note is not published.
Expand All @@ -38,7 +38,7 @@ what each shorthand denotes and where (if anywhere) it can be obtained.

## Why these are not in the repository

The proposal and the archived note are internal planning records; the
The proposal and the internal note are internal planning records; the
manuscript is a separate (in-preparation) publication. Publishing them here
would conflate planning history with normative specification. The calculus is
stated so that its normative content does not depend on reading either.
Expand Down Expand Up @@ -78,7 +78,7 @@ conditional cell or as a recorded input, and the consumer never validates it.
## Historical provenance

Shorthand citations (`proposal §N`, `code-deformation`, `manuscript`,
`an archived internal note`) are historical provenance only. The normative
`an internal note`) are historical provenance only. The normative
content of `spec/` does not depend on reading any of them, and they are not
required to understand or use the calculus. Where a result above is reproduced,
the reproduction in `spec/OPERATIONAL_SEMANTICS.md` §4 is the authoritative
Expand Down
6 changes: 3 additions & 3 deletions lean/CodeCalculus/Soundness.lean
Original file line number Diff line number Diff line change
Expand Up @@ -30,7 +30,7 @@ The pieces, in dependency order:
- `Chain` — the chain of states a closed command unfolds through; this is the
"same closed term" that `TRANSITIVITY` is derivation-wide over.
- `Derives` — `Γ ⊢ … : R`, the L3 rule inventory of
`the archived notes` §3.4 and §3.4.1:
`the internal notes` §3.4 and §3.4.1:
instantiation, `TRANSITIVITY`, `MIN-ORDER`, and the ambient order theory.
- `Models`, `judgment_soundness` — the models predicate, and the metatheorem.

Expand Down Expand Up @@ -88,7 +88,7 @@ of atoms at evaluation sites, but it would also add premises to those two rules;
the absorbing encoding keeps the rules as §3.4.1 states them. A consequence,
recorded rather than repaired: a sentence true over a chain can become false
over an extension of it when a `min` argument that did not denote begins to
(`an archived note`, `rg_optMin_breaks_chain_monotonicity`); the transport
(`an internal note`, `rg_optMin_breaks_chain_monotonicity`); the transport
lemmas of `Transport.lean` are stated on `denotes`, not on values, for that
reason. Second, `sub` is strict in `none` and is used by the graft's `n - 1`
schema and by the truncated readings of the two `k` cells; `add` is its exact
Expand Down Expand Up @@ -457,7 +457,7 @@ theorem measureMonotone_rowCount : MeasureMonotone (fun _ s => s.rows.length) wh

/-- Derivability of an annotation in the derivation of the chain as one closed
term: the L3 rule inventory of
`the archived notes` §3.4/§3.4.1.
`the internal notes` §3.4/§3.4.1.

The `inst*` rows are §3.4's instantiation — "at every evaluation of an
operation, each claim schema that the operation's card records as `included`
Expand Down
4 changes: 2 additions & 2 deletions machine/certificates.py
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
"""Certificate format and checker prototype for sector distance lower bounds.

This module is the first branch of the certificate layer explored in
`the archived notes` (option (ii), opened 2026-10-25). It makes the *lower* endpoint of a sector distance checkable the
`the internal notes` (option (ii), opened 2026-10-25). It makes the *lower* endpoint of a sector distance checkable the
same way the upper endpoint already is: an untrusted producer emits a
self-contained artifact, and a small total checker accepts or rejects it
without searching.
Expand All @@ -23,7 +23,7 @@
``d_sigma >= L`` holds. The checker is the trusted base; the producer is
arbitrary and untrusted, exactly as for the ledger's logical witnesses.

Checker contract (`the archived notes` section 4):
Checker contract (`the internal notes` section 4):

1. Total and decidable: a pure predicate over a finite input; it returns
accept/reject with reasons and never raises on malformed input.
Expand Down
8 changes: 4 additions & 4 deletions machine/experiments/certificate_prototype.json
Original file line number Diff line number Diff line change
Expand Up @@ -3,10 +3,10 @@
"format": "css-sector-distance-lower-bound",
"format_version": 1,
"instance": {
"source": "unitaryfoundation/qldpc-challenge an archived note",
"source": "unitaryfoundation/qldpc-challenge an internal note",
"pinned_commit": "9d30c66a43bf7c275196267d369ddab171bfd72f",
"fixture": "machine/fixtures/qldpc_10_2_3_b.json",
"fixture_sha256": "84a5366a226544f11fdaae2bd4c434702ceaacf43022df85c0c47b656652fe6f",
"fixture_sha256": "e359d01afe8083d53df1e788d4f42165f62271e3f5fd744e94daddf89df3107c",
"n": 10,
"k": 2
},
Expand Down Expand Up @@ -37,11 +37,11 @@
},
"checker": {
"module": "machine/certificates.py",
"contract": "an archived note section 4",
"contract": "an internal note section 4",
"coverage_enumeration_per_certificate": 56
},
"kill_criterion": {
"statement": "an archived note section 8",
"statement": "an internal note section 8",
"verdict": "passes at this instance",
"assessment": "The prototype certifies a non-trivial lower endpoint (d_X >= 3 and d_Z >= 3 on the pinned parent) with 56 kill entries and a few kilobytes per sector, and the checker is total and decidable over a small trusted base. The flat ball format's cost is sum(C(n, w), w < bound) entries; at the challenge's larger boards the same format is impractically large (see scaling_probes), so the next decision is the plan's section 9 question 1 -- search tree or covering certificate -- not a bigger flat artifact.",
"scaling_probes": [
Expand Down
10 changes: 5 additions & 5 deletions machine/experiments/certificate_prototype.py
Original file line number Diff line number Diff line change
Expand Up @@ -5,10 +5,10 @@
by post-hoc exhaustive enumeration -- an oracle the consumer never calls.
This script turns the *lower-bound half* of that oracle into a checkable
artifact: a kill certificate per sector for ``d_sigma >= 3`` on the pinned
``an archived note`` parent, verified by the checker in
``an internal note`` parent, verified by the checker in
`machine/certificates.py`. It records the artifact sizes and the
kill-criterion assessment required by the first-branch deliverable
(`an archived note` sections 5 and 8).
(`an internal note` sections 5 and 8).

The upper-bound half stays where it already was: a logical witness the
ledger validates by finite-field arithmetic. Nothing here changes a claim
Expand Down Expand Up @@ -101,7 +101,7 @@ def build_prototype_artifact():
"format": "css-sector-distance-lower-bound",
"format_version": 1,
"instance": {
"source": "unitaryfoundation/qldpc-challenge an archived note",
"source": "unitaryfoundation/qldpc-challenge an internal note",
"pinned_commit": PINNED_COMMIT,
"fixture": "machine/fixtures/qldpc_10_2_3_b.json",
"fixture_sha256": hashlib.sha256(fixture_bytes).hexdigest(),
Expand All @@ -112,12 +112,12 @@ def build_prototype_artifact():
"negative_control": negative_control,
"checker": {
"module": "machine/certificates.py",
"contract": "an archived note section 4",
"contract": "an internal note section 4",
"coverage_enumeration_per_certificate":
_ball_entries(state.n(), CERTIFIED_BOUND),
},
"kill_criterion": {
"statement": "an archived note section 8",
"statement": "an internal note section 8",
"verdict": "passes at this instance",
"assessment": (
"The prototype certifies a non-trivial lower endpoint "
Expand Down
12 changes: 6 additions & 6 deletions machine/experiments/challenge_replay.json
Original file line number Diff line number Diff line change
@@ -1,7 +1,7 @@
{
"qldpc_commit": "9d30c66a43bf7c275196267d369ddab171bfd72f",
"addition": {
"parent": "an archived note",
"parent": "an internal note",
"parent_parameters": {
"n": 10,
"k": 2
Expand Down Expand Up @@ -143,7 +143,7 @@
"interpretation": "The propagation path used the parent's exhaustive lower bound, C-ADD-CHECK monotonicity, and child logical witnesses. The second exact enumeration was an independent experiment-only check, not part of the consumer path."
},
"strict_improvement_survey": {
"parent": "an archived note",
"parent": "an internal note",
"parent_exact_distance_experiment": 3,
"support_vectors_examined_per_sector": 1023,
"exact_distance_oracle": "experiment-only; not called by the consumer; minimum weight over all logicals in each sector",
Expand All @@ -163,7 +163,7 @@
}
},
"graft": {
"parent": "an archived note",
"parent": "an internal note",
"operation": {
"op": "E-GRAFT",
"qubit": 0,
Expand Down Expand Up @@ -291,10 +291,10 @@
"interpretation": "The consumer prunes target 3 from a valid weight-2 upper witness. Post-hoc exhaustive enumeration finds the child's distance to be 2, so the transported upper bound is exact on this instance, and the spliced sector is unchanged at 3 as the source's sector-wise relation predicts. The randomized verifier's failure to find a lighter logical is consistent with that, and is still not a lower-bound certificate: it is a randomized refutation, and the ledger's lower endpoint remains 0."
},
"deletion": {
"parent": "an archived note",
"parent": "an internal note",
"children": [
"an archived note",
"an archived note"
"an internal note",
"an internal note"
],
"final_child_parameters": {
"n": 119,
Expand Down
30 changes: 15 additions & 15 deletions machine/experiments/challenge_replay.py
Original file line number Diff line number Diff line change
Expand Up @@ -41,7 +41,7 @@ def exact_distances(state):
* It returns the **minimum weight** logical in each sector. The scan
is exhaustive, so the first logical met in integer order is not
necessarily the lightest: integer order is not weight order. On the
pinned ``an archived note`` the first-integer-order reading gives
pinned ``an internal note`` the first-integer-order reading gives
5 where the true distance is 3.
* It enumerates over the state's own qubit **labels**. Row masks are
indexed by label, not by position, so a state whose namespace is not
Expand Down Expand Up @@ -111,7 +111,7 @@ def find_witness(state, side, max_weight=3):
def _run_challenge_verifier(challenge_root, document, directory, label):
path = directory / (label + ".json")
path.write_text(json.dumps(document, indent=2) + "\n")
verifier = Path(challenge_root) / "verify" / "an archived note"
verifier = Path(challenge_root) / "verify" / "an internal note"
result = subprocess.run(
[sys.executable, str(verifier), str(path)],
check=False,
Expand Down Expand Up @@ -141,7 +141,7 @@ def _run_challenge_verifier(challenge_root, document, directory, label):


def run_addition_replay(challenge_root, output_dir):
parent_path = Path(challenge_root) / "codes" / "an archived note"
parent_path = Path(challenge_root) / "codes" / "an internal note"
parent = json.loads(parent_path.read_text())
initial_state = CodeState.from_challenge(parent)
measured_parent = exact_distances(initial_state)
Expand All @@ -168,12 +168,12 @@ def run_addition_replay(challenge_root, output_dir):
output = export_challenge_entry(
parent, final_state, ledger,
"[[10,1,3]] priced-calculus addition",
"C-ADD-CHECK X {0,8,9} from challenge an archived note",
"C-ADD-CHECK X {0,8,9} from challenge an internal note",
"2026-10-19",
)
verifier = _run_challenge_verifier(challenge_root, output, output_dir, "addition")
return {
"parent": "an archived note",
"parent": "an internal note",
"parent_parameters": {"n": initial_state.n(), "k": initial_state.k()},
"parent_exact_distance_experiment": measured_parent,
"operation": operation,
Expand All @@ -194,13 +194,13 @@ def run_addition_replay(challenge_root, output_dir):
def run_strict_improvement_survey(challenge_root):
"""Exhaust every single-check addition to the small challenge parent."""
parent = json.loads(
(Path(challenge_root) / "codes" / "an archived note").read_text()
(Path(challenge_root) / "codes" / "an internal note").read_text()
)
initial = CodeState.from_challenge(parent)
old_distance = exact_distances(initial)["d"]
support_vectors = (1 << initial.n()) - 1
report = {
"parent": "an archived note",
"parent": "an internal note",
"parent_exact_distance_experiment": old_distance,
"support_vectors_examined_per_sector": support_vectors,
"exact_distance_oracle": "experiment-only; not called by the consumer; "
Expand Down Expand Up @@ -234,9 +234,9 @@ def run_strict_improvement_survey(challenge_root):


def run_deletion_replay(challenge_root, output_dir):
parent = json.loads((Path(challenge_root) / "codes" / "an archived note").read_text())
child1 = json.loads((Path(challenge_root) / "codes" / "an archived note").read_text())
child2 = json.loads((Path(challenge_root) / "codes" / "an archived note").read_text())
parent = json.loads((Path(challenge_root) / "codes" / "an internal note").read_text())
child1 = json.loads((Path(challenge_root) / "codes" / "an internal note").read_text())
child2 = json.loads((Path(challenge_root) / "codes" / "an internal note").read_text())
operations = [
{"op": "C-DELETE-CHECK", "row_id": "a1",
"witnesses": {"X": child1["distance"]["X"]["witness"],
Expand All @@ -262,8 +262,8 @@ def run_deletion_replay(challenge_root, output_dir):
verifier = _run_challenge_verifier(challenge_root, directory=output_dir,
document=output, label="deletion")
return {
"parent": "an archived note",
"children": ["an archived note", "an archived note"],
"parent": "an internal note",
"children": ["an internal note", "an internal note"],
"final_child_parameters": {"n": state.n(), "k": state.k()},
"consumer_trace": trace,
"final_interval": ledger.as_json()["d"],
Expand All @@ -280,7 +280,7 @@ def run_deletion_replay(challenge_root, output_dir):

def run_graft_replay(challenge_root, output_dir):
parent = json.loads(
(Path(challenge_root) / "codes" / "an archived note").read_text()
(Path(challenge_root) / "codes" / "an internal note").read_text()
)
initial_state = CodeState.from_challenge(parent)
operation = {
Expand Down Expand Up @@ -324,14 +324,14 @@ def run_graft_replay(challenge_root, output_dir):
output = export_challenge_entry(
parent, state, ledger,
"[n=9,k=2] E-GRAFT witness submission",
"E-GRAFT q=0, X chain [a1,a2] from an archived note",
"E-GRAFT q=0, X chain [a1,a2] from an internal note",
"2026-10-19",
)
verifier = _run_challenge_verifier(
challenge_root, output, output_dir, "graft"
)
return {
"parent": "an archived note",
"parent": "an internal note",
"operation": operation,
"child_parameters": {"n": state.n(), "k": state.k()},
"bounded_witness_search_candidates": bounded_search,
Expand Down
2 changes: 1 addition & 1 deletion machine/experiments/priced_conformance_harness.py
Original file line number Diff line number Diff line change
Expand Up @@ -2,7 +2,7 @@

This harness is the instrument named in . It generates small random
CSS states, applies every admitted transition, measures the exact per-sector
distances with the external oracle in ``an archived note``, and
distances with the external oracle in ``an internal note``, and
asserts that every interval emitted by ``machine.priced_calculus`` contains
the corresponding exact distance.

Expand Down
2 changes: 1 addition & 1 deletion machine/fixtures/qldpc_10_2_3_b.json
Original file line number Diff line number Diff line change
Expand Up @@ -87,7 +87,7 @@
"origin": "submission",
"date": "2026-09-27",
"model": "Mimo-V2.6-Free",
"notes": "Equivalence assessed against the board: no exact stabilizer-group fingerprint match and no Weisfeiler-Leman signature match to any existing entry. Filed as `an archived note`, a lighter-check companion to `an archived note` (same [[10,2,3]], max check weight 4 vs 6)."
"notes": "Equivalence assessed against the board: no exact stabilizer-group fingerprint match and no Weisfeiler-Leman signature match to any existing entry. Filed as `an internal note`, a lighter-check companion to `an internal note` (same [[10,2,3]], max check weight 4 vs 6)."
},
"family": "topological",
"locality": {
Expand Down
2 changes: 1 addition & 1 deletion machine/lean_fixture_coverage.py
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@
- a mismatch count (uncovered fixtures).

What it cannot see is stated in the report and in
`the archived notes`. Since the Lean side
`the internal notes`. Since the Lean side
replays the single-step and two-step fixtures under `Eval` and `EvalExpr`
(`lean/CodeCalculus/FixtureEvalReplay.lean`); this report counts
well-formedness theorems only and does not read those replays. Distances and
Expand Down
8 changes: 4 additions & 4 deletions machine/planner.py
Original file line number Diff line number Diff line change
Expand Up @@ -6,15 +6,15 @@
and classifying the result as ``certify`` / ``prune`` / ``re-measure`` against
a target. It never applies a move, never recomputes a distance, price, or
witness, and never discharges an obligation: those remain the consumer's and
the calculus's respectively (see the archived notes).
the calculus's respectively (see the internal notes).

First milestone: deletion chains (the move that actually pays) plus the
other single-qubit-removing moves (peel, puncture). Search control (beam
width, depth, memoisation) is consumer configuration and lives in the
caller.

adds the two cheap screens the deletion campaign used before its
witness search (qldpc-challenge `an archived note`: 28,913 of 88,592
witness search (qldpc-challenge `an internal note`: 28,913 of 88,592
essential single-check deletions -- 33% -- discarded as freeing a bare
qubit, "a structural filter on column weights"; 40 gate-passing candidates
reduced to 18 by domination before filing):
Expand All @@ -26,7 +26,7 @@
on every axis the ledger exports by another survivor.

Both are consumer configuration, not calculus data; the calculus's
transport rules are untouched (see the archived notes).
transport rules are untouched (see the internal notes).
"""


Expand Down Expand Up @@ -96,7 +96,7 @@ def has_bare_qubit(state):
1; k stays >= 1 because both sectors' row spaces are contained in the
codimension-1 subspace of vectors that are zero at q, and their
dimension sum is at most n-1 there (the argument is in
the archived notes). No witness search can put the state
the internal notes). No witness search can put the state
above distance 1, which is what makes the screen cheap and safe.
This is a fact about the state the consumer screens on, not a
calculus claim: the transport rules of `machine/priced_calculus.py`
Expand Down
2 changes: 1 addition & 1 deletion machine/state.json
Original file line number Diff line number Diff line change
Expand Up @@ -7,7 +7,7 @@
],
"authority": "This file is a faithful transcription for mechanization consumers. Where it disagrees with spec/, the prose wins and this file must be fixed.",
"generated": "2026-10-27",
"representation": "sparse supports over qubit indices; the encoding qldpc-challenge already ships an archived note for its own codes, which this file does not transcribe."
"representation": "sparse supports over qubit indices; the encoding qldpc-challenge already ships an internal note for its own codes, which this file does not transcribe."
},
"state": {
"Q": {
Expand Down
2 changes: 1 addition & 1 deletion machine/test_pinned_replay.py
Original file line number Diff line number Diff line change
Expand Up @@ -6,7 +6,7 @@

HERE = os.path.dirname(os.path.abspath(__file__))

PINNED_SHA256 = "d9a48e3116df21987b953c146f8b15ddefe9d1a05ad8a3bcb070801360827539"
PINNED_SHA256 = "5078db2ff2a447b9a0f4ea387b8ab28bae0be804daaff1bee75b0a8fc8b275e3"
REPLAY_PATH = os.path.join(HERE, "experiments", "challenge_replay.json")


Expand Down
2 changes: 1 addition & 1 deletion machine/test_priced_calculus.py
Original file line number Diff line number Diff line change
Expand Up @@ -598,7 +598,7 @@ class LayoutTransportTests(unittest.TestCase):

The interaction radius is a function of the check set, so a parent's
claim cannot be carried into a child whose checks differ. See
`the archived notes` and .
`the internal notes` and .
"""

def test_radius_is_the_largest_check_diameter(self):
Expand Down
Loading
Loading