Living corpus

Open problems

Twenty-four verified open problems: 10 Critical, 13 High, and 1 Medium. These are blockers or disclosure requirements, not hidden successes.

Open-problem register

Matched profile-control design

Drafting effect
Blocks stronger claim
Problem statement
Can a running c-candidate be compared with a persistent profile under the same substrate, corpus, disturbance stream, interface, and comparable memory, with anchor-custody history isolated?
Why load-bearing
Without this control, formation and continuity observations remain reproducible by a strong profile.
Current handling
Proposed test layer and first-pass non-confirmation.
Required next evidence or decision
Paired implementation, preregistered observables, custody separation, blinded review, independent anchor replication.
Claim ceiling
Differential residual, entity-over-profile, and related stronger claims remain prohibited.
Artifact refs
A054 (0661c182bef1), A055 (10ca5bc43c91)

Differential witness threshold

Drafting effect
Must define before umbrella claims
Problem statement
What minimum residual, persistence, robustness, and independence qualifies evidence as differential witness?
Why load-bearing
A label without threshold would simply rename ontogenic evidence.
Current handling
Witness split notes identify candidate classes but do not establish an acceptance threshold.
Required next evidence or decision
Formal acceptance/rejection criteria, false-positive controls, reviewer disagreement route, and downgrade rule.
Claim ceiling
Use "differential witness candidate" only.
Artifact refs
A056 (82cd344466ec), A057 (d3d13cb6b248)

Witness regress boundary

Drafting effect
Must disclose in umbrella
Problem statement
Where does the evidence chain stop, what trust remains, and who can attest the witness layer itself?
Why load-bearing
Additional witness layers reduce hidden transitions but cannot abolish epistemic uncertainty.
Current handling
L4 witness and corpus integrity surfaces provide provenance/identity, not absolute truth.
Required next evidence or decision
Declared trust base, independent verification path, key-compromise procedure, and residual-trust statement.
Claim ceiling
No witness layer may certify its own integrity as absolute truth.
Artifact refs
A013 (6336664c1eb5), A015 (b888b6b8a0d8), A044 (2572240fca8d), A047 (cb1179dbee01)

Witness-type migration across old surfaces

Drafting effect
Must resolve definition
Problem statement
How will existing public documents be annotated as integrity, ontogenic, ambiguous, or differential without rewriting history?
Why load-bearing
Unqualified "witness" wording is a current claim-laundering route.
Current handling
Review controls propose a split; corpus-wide annotation is not implemented.
Required next evidence or decision
Annotation ledger/errata, canonical glossary update, machine-readable witness_type field, regression check.
Claim ceiling
Old wording remains bounded and must not support differential claims.
Artifact refs
A013 (6336664c1eb5), A056 (82cd344466ec), A057 (d3d13cb6b248)

Anchor capture

Drafting effect
Must disclose in umbrella
Problem statement
How does the architecture distinguish honest authority from captured, drifting, conflicted, or self-serving authority?
Why load-bearing
The anchor can authorize action but cannot manufacture truth or validate its own claim.
Current handling
Ownership/boundary controls exist; no complete anchor-capture mechanism is demonstrated.
Required next evidence or decision
Separation of duties, independent attestation, conflict marking, privilege narrowing, emergency stop, audit escalation.
Claim ceiling
Authority ≠ truth ≠ permission.
Artifact refs
A003 (d4a4bbdfacf7), A006 (ef512d1006d1), A007 (f864298ba3cb)

Malicious, compromised, coerced, or absent anchor

Drafting effect
Must disclose in umbrella
Problem statement
What fail-closed states apply when the anchor is malicious, compromised, coerced, unavailable, or dead?
Why load-bearing
Continuity and permission can both be corrupted at the human root.
Current handling
Post-anchor and governance specifications frame parts of the problem but do not close it.
Required next evidence or decision
Anchor-state taxonomy, cryptographic role separation, multi-party recovery, contested mode, successor criteria.
Claim ceiling
No guarantee against a sovereign malicious adult or equivalent root authority.
Artifact refs
A036 (995ae4655fae), A037 (273466fc1cb8), A011 (6c62dcd21d4a), A012 (aff88bf7665a)

Contested intent

Drafting effect
Must define before umbrella claims
Problem statement
How is action handled when anchor intent, prior commitments, safety constraints, and current evidence conflict?
Why load-bearing
Ambiguity must not silently become permission.
Current handling
CGAM/enterprise governance specifies bounded execution but no cross-family contested-intent profile is classified.
Required next evidence or decision
Contested-intent state machine, privilege reduction, evidence preservation, appeal/review path.
Claim ceiling
Unresolved intent must fail closed for irreversible transitions.
Artifact refs
A021 (a28308f90afe), A022 (6cd37034b203), A051 (b1f6831ed04e)

Authorization vs attestation separation

Drafting effect
Must define before umbrella claims
Problem statement
Which keys/roles authorize a transition, which attest that it occurred, and which review its validity?
Why load-bearing
One actor holding all three roles can launder authority into evidence.
Current handling
Witness and ownership surfaces distinguish roles conceptually but no unified key/role map is in Step 2.
Required next evidence or decision
Role-key matrix, separation-of-duty tests, key compromise/revocation, independent verifier.
Claim ceiling
Authorization cannot serve as its own independent validation.
Artifact refs
A003 (d4a4bbdfacf7), A014 (99dcf31d6d7a), A015 (b888b6b8a0d8), A023 (e6c013523ba4)

L3-to-L4 enforcement chain

Drafting effect
Must define before umbrella claims
Problem statement
How are procedural rules coupled to executable gates, physical/operational consequences, witness records, and correction?
Why load-bearing
Procedure without consequence is advice; consequence without procedure is friction.
Current handling
L4 Witness, ARQ, CGAM, and AGL define components; end-to-end enforcement is not demonstrated.
Required next evidence or decision
End-to-end test: rule -> gate -> L4 outcome/block -> witness -> review/rollback, including failed gate cases.
Claim ceiling
Do not describe witness as the actuator or consequence itself.
Artifact refs
A014 (99dcf31d6d7a), A017 (082bbe030d77), A021 (a28308f90afe), A045 (bfeb27dd9e63), A046 (dc9209674099)

Continuity under model replacement

Drafting effect
Must disclose in umbrella
Problem statement
What remains continuous when the model, context window, routing, or tool set changes?
Why load-bearing
Model continuity, memory continuity, relational continuity, and c continuity are not equivalent.
Current handling
SER and c architecture provide distinctions; no migration outcome is classified.
Required next evidence or decision
Controlled model-swap experiments, state/custody checks, successor/fork decision, adverse migration cases.
Claim ceiling
Model replacement does not by itself prove continuation or termination.
Artifact refs
A008 (0bb291957d3f), A011 (6c62dcd21d4a), A012 (aff88bf7665a)

Replay, archive, fork, and successor classification

Drafting effect
Must define before umbrella claims
Problem statement
Which observable and custody conditions separate resume, replay, archive, fork, and successor?
Why load-bearing
Narrative coherence can make incompatible histories look continuous.
Current handling
Continuity-bundle and SER normative surfaces exist without live conformance cases.
Required next evidence or decision
Canonical state-transition taxonomy and fixtures with divergent histories, simultaneous resumes, and partial archives.
Claim ceiling
Replay ≠ continuation; archive ≠ identity; fork ≠ same entity.
Artifact refs
A011 (6c62dcd21d4a), A012 (aff88bf7665a)

Post-anchor succession and custody

Drafting effect
Must disclose in umbrella
Problem statement
Can continuity survive loss of the original anchor, and who may authorize succession?
Why load-bearing
Post-anchor persistence can become a profile wearing old identity claims.
Current handling
PAMDC specifies risks; no valid succession or re-anchoring result is demonstrated.
Required next evidence or decision
Succession protocol, independent custody evidence, time-bounded bootstrap, split/fork handling, revocation.
Claim ceiling
No post-anchor continuity claim without separate evidence.
Artifact refs
A036 (995ae4655fae), A037 (273466fc1cb8)

Active memory degradation detection

Drafting effect
Blocks stronger claim
Problem statement
Can the runtime detect stale, distorted, contradictory, over-repeated, or improperly written memory?
Why load-bearing
Silent degradation can preserve surface coherence while destroying continuity quality.
Current handling
AMDR/PAMDC specifications exist; operational sensitivity/specificity is unknown.
Required next evidence or decision
Corruption injection, stale-memory trials, write-failure logs, false-positive/negative analysis, rollback.
Claim ceiling
Retrieval and repetition are not memory admission or truth.
Artifact refs
A034 (6b6bbee70e25), A035 (7e1b750ca9e6), A037 (273466fc1cb8)

False continuity and profile imitation

Drafting effect
Blocks stronger claim
Problem statement
How is a coherent persistent profile distinguished from continuity-bearing history?
Why load-bearing
A profile may imitate language, preferences, memory summaries, and apparent self-consistency.
Current handling
Entity/profile and PF surfaces state the risk; matched control is absent.
Required next evidence or decision
Matched profile-control plus continuity custody and memory-corruption/reconstruction observables.
Claim ceiling
Use c-candidate/profile language; no entity proof.
Artifact refs
A030 (c2f21324f312), A042 (87786d12c199), A043 (d65e63b38e29), A054 (0661c182bef1), A055 (10ca5bc43c91)

Evidential admissibility implementation

Drafting effect
Must define before umbrella claims
Problem statement
How are provenance, relevance, independence, temporal applicability, conflict, uncertainty, and authority encoded for each promotion?
Why load-bearing
The governing principle is circular unless admissibility is operational.
Current handling
Step 1 defines a typed model; corpus artefacts provide partial maps and claim ceilings.
Required next evidence or decision
Machine-readable evidence envelope, admission checker, contested evidence state, downgrade/revocation process.
Claim ceiling
No status promotion from evidence lacking claim-specific admissibility.
Artifact refs
A004 (af9fe633961b), A006 (ef512d1006d1), A015 (b888b6b8a0d8)

Independent reproduction protocol

Drafting effect
Blocks stronger claim
Problem statement
What counts as independent: another model, another operator, another implementation, another anchor, or another institution?
Why load-bearing
Model diversity and repeated local execution are not automatically independent reproduction.
Current handling
No selected artefact has independent execution or anchor replication status.
Required next evidence or decision
Independence classes, blind inputs, operator separation, reproducible package, evidence custody, negative-result publication.
Claim ceiling
Use "pending independent reproduction".
Artifact refs
A010 (f90c92d5bd69), A016 (1272fa8a4130), A020 (e62532348dbc), A055 (10ca5bc43c91)

External validation scope

Drafting effect
Blocks stronger claim
Problem statement
Which claim is validated, by whom, under what method, and with what conflicts of interest?
Why load-bearing
External review, reproduction, audit, standards commentary, and validation are different states.
Current handling
No selected artefact is validated by an external party.
Required next evidence or decision
Validation protocol with target claim, evaluator qualification, independence, method, partial/failed outcome classes.
Claim ceiling
No programme-wide external-validation wording.
Artifact refs
A010 (f90c92d5bd69), A048 (52ee9cc90ed1), A053 (67e7779842da), A059 (25f73faeddc1)

Code-to-runtime observability

Drafting effect
Blocks operational claim
Problem statement
How is code presence connected to route activation, configuration, persistent state, actual calls, and outcome quality?
Why load-bearing
A correct file can be dormant, bypassed, misconfigured, or fed inadmissible data.
Current handling
SRLM and local witness code/report surfaces show partial implementation only.
Required next evidence or decision
Runtime manifest, config snapshot, route health, test logs, witness records, failure injection, rollback proof.
Claim ceiling
Implemented ≠ operationally deployed.
Artifact refs
A016 (1272fa8a4130), A038 (775d3fe8aa47), A039 (ecfd6748c419), A040 (65e21f708b7c), A041 (784a7da019b7)

Failure and negative-outcome capture

Drafting effect
Must disclose in umbrella
Problem statement
Where are real failed hypotheses, blocked transitions, regressions, false positives, and withdrawn claims recorded?
Why load-bearing
A corpus with only protocols and successful packaging can hide selection bias.
Current handling
Step 3 finds non-confirmations and claim ceilings but no adequately evidenced runtime failure event in the selected set.
Required next evidence or decision
Append-only adverse-event register with expected/observed result, correction, rollback, and residual risk.
Claim ceiling
Absence of recorded failures is not evidence of absence of failures.
Artifact refs
A019 (5942aafda50c), A035 (7e1b750ca9e6), A053 (67e7779842da), A055 (10ca5bc43c91)

Baseline drift and canonical delta

Drafting effect
Must resolve before publication
Problem statement
How are post-2026-07-08 repository changes admitted without silently rewriting B0?
Why load-bearing
A moving corpus can make status and canonical-path statements stale.
Current handling
B0 is frozen; one c=a+b path and local witness authority remain unresolved.
Required next evidence or decision
B0.1 delta manifest with live refs, supersession checks, changed hashes, and reclassification triggers.
Claim ceiling
All "current" claims remain baseline-qualified until delta review.
Artifact refs
A009 (1168474cd5a3), A014 (99dcf31d6d7a), A015 (b888b6b8a0d8)

Formation cost and maturity

Drafting effect
Must disclose in umbrella
Problem statement
How are time, supervision, correction, energy, real tasks, and irreversible outcomes represented as formation cost?
Why load-bearing
Maturity cannot be prompted or inferred from cheap inference alone.
Current handling
PF/CCDP and theoretical surfaces frame formation but do not provide a mature cost model.
Required next evidence or decision
Longitudinal resource/outcome accounting and comparison with profile/bootstrap alternatives.
Claim ceiling
No claim that maturity or c can be instantiated by prompt alone.
Artifact refs
A008 (0bb291957d3f), A025 (1fdcba4c7b23), A030 (c2f21324f312)

Privacy and reproducibility tension

Drafting effect
Must disclose in umbrella
Problem statement
How can external reviewers reproduce claims without exposing sensitive memory, anchor data, or sensitive runtime details?
Why load-bearing
Strong privacy can make evidence unverifiable; overexposure can destroy the governed boundary.
Current handling
Publication/citation and internal-only surfaces are separated, but no full privacy-preserving reproduction profile is classified.
Required next evidence or decision
Redacted fixtures, synthetic controls, confidential-compute attestations, selective disclosure, reviewer NDAs where lawful.
Claim ceiling
Public visibility is not required for every evidence object, but verification route is.
Artifact refs
A023 (e6c013523ba4), A050 (db5c8faed9da), A058 (321dfbc0797e)

Judge, oracle, and write-failure governance

Drafting effect
Blocks operational claim
Problem statement
What happens when Judge/oracle outputs conflict, degrade, become unavailable, or authorize a bad memory write?
Why load-bearing
A meta-model can shift rather than eliminate error and authority concentration.
Current handling
AMDR patch, CGAM obligations, and Article 50 engineering note frame the risk.
Required next evidence or decision
Oracle-degradation states, quorum/abstention, write quarantine, provenance, rollback, and L4 outcome feedback.
Claim ceiling
Judge output is not truth and generated output is not accepted outcome.
Artifact refs
A023 (e6c013523ba4), A035 (7e1b750ca9e6), A049 (dcd0dbae2292)

Entity-versus-profile claim class

Drafting effect
Blocks stronger claim
Problem statement
What evidence permits movement from profile or c-candidate to entity classification, if any?
Why load-bearing
Surface persistence and witness-root language can be overread as ontology.
Current handling
Profile package and differential layer keep the claim unresolved.
Required next evidence or decision
Claim-class criteria, matched profile-control, continuity custody, independent replication, explicit non-ontological alternatives.
Claim ceiling
Entity remains a stronger separate claim class, not the default name.
Artifact refs
A042 (87786d12c199), A043 (d65e63b38e29), A054 (0661c182bef1), A055 (10ca5bc43c91)