Parent handoff milestone: #721
Related: #605 #657 #697–#719 #525 #691
Editorial problem
The late v2.10+ issue family correctly specifies many scientific requirements for learned-model research, protected evaluation, score-to-action policy, deployment, retraining, experiment identity and attribution. After freezing the public/private boundary in #721, those requirements need an explicit ownership classification so histdata.com-tools does not accidentally become the operational learned-model repository.
The public project should retain reusable scientific/data interoperability contracts where they are required to preserve provenance, causality or compatibility. Concrete proprietary model research/execution should move to the separate private ML repository.
Without this editorial pass, issue wording such as build, train, deploy, champion/challenger, live, or retrain can be interpreted as requiring public implementation of model registries, trained weights and operational deployment state before #721 may close.
Outcome
Review every active learned-model/decision/deployment issue in the late #605/#657 lineage, especially #696–#719, and classify each requirement atomically into one of four ownership states:
public_required_contract — schema, causal rule, validation fixture or compatibility requirement required for the published substrate/handoff;public_reference_optional — neutral reference implementation/test useful publicly but not a handoff blocker unless another public product consumes it;private_execution_required — concrete model fitting, hyperparameter search, trained weights, fitted preprocessing/decision policies, protected result ledgers, deployment state or retraining operations owned by the private ML repository;deferred_cross_repo_interface — shared contract whose public/private implementations are both future work and not required to certify the initial handoff.
Required editorial review
At minimum review issue families covering:
- labels/target contracts;
- train-only preprocessing;
- protected model-selection/search governance;
- domain/regime evaluation;
- instrument/session/universe semantics;
- financing, margin and execution-cost semantics where they affect public synthetic labels/features;
- uncertainty propagation;
- historical-to-modern transport;
- offline/online parity and live-validity concepts;
- score-to-action policy;
- deployment/rollback;
- champion/challenger/retraining;
- experiment bundles;
- attribution/explanation artifacts.
The same issue may contain both public and private atoms. Split or rewrite it when one status cannot accurately describe both.
Governing boundary
Public histdata.com-tools owns the scientific substrate and neutral contracts needed to describe or verify downstream usage. It does not own concrete proprietary:
- model architecture/search campaigns beyond public reference/forecast models already part of the upstream data product;
- trained trading-model weights;
- fitted experiment-specific preprocessing;
- private target variants;
- hyperparameter/search history;
- protected performance results;
- fitted score-to-action thresholds/sizing;
- live deployment registry/state;
- champion/challenger promotion state;
- retraining schedules/state;
- model-specific production attribution archives.
The autonomous economic-release forecasting factory #556 is an upstream public data-producing subsystem and is therefore not reclassified merely because it uses ML internally. This issue concerns downstream learned trading/decision research over the final substrate.
Issue mutation rules
For every reviewed issue:
Handoff manifest
#721's handoff bundle must expose the public contract versions that downstream private implementations are expected to reference. It must not embed private model artifacts.
Acceptance
Every late ML issue has machine-readable ownership at requirement granularity; #721's dependency graph contains only public responsibilities; private execution work remains traceable for the new repository but cannot prevent the public data/simulation substrate from reaching terminal handoff; no learned-model operational state is silently required to reside in histdata.com-tools; and #691 proves zero orphaned requirements after the editorial split.
Parent handoff milestone: #721
Related: #605 #657 #697–#719 #525 #691
Editorial problem
The late v2.10+ issue family correctly specifies many scientific requirements for learned-model research, protected evaluation, score-to-action policy, deployment, retraining, experiment identity and attribution. After freezing the public/private boundary in #721, those requirements need an explicit ownership classification so
histdata.com-toolsdoes not accidentally become the operational learned-model repository.The public project should retain reusable scientific/data interoperability contracts where they are required to preserve provenance, causality or compatibility. Concrete proprietary model research/execution should move to the separate private ML repository.
Without this editorial pass, issue wording such as
build,train,deploy,champion/challenger,live, orretraincan be interpreted as requiring public implementation of model registries, trained weights and operational deployment state before #721 may close.Outcome
Review every active learned-model/decision/deployment issue in the late #605/#657 lineage, especially #696–#719, and classify each requirement atomically into one of four ownership states:
public_required_contract— schema, causal rule, validation fixture or compatibility requirement required for the published substrate/handoff;public_reference_optional— neutral reference implementation/test useful publicly but not a handoff blocker unless another public product consumes it;private_execution_required— concrete model fitting, hyperparameter search, trained weights, fitted preprocessing/decision policies, protected result ledgers, deployment state or retraining operations owned by the private ML repository;deferred_cross_repo_interface— shared contract whose public/private implementations are both future work and not required to certify the initial handoff.Required editorial review
At minimum review issue families covering:
The same issue may contain both public and private atoms. Split or rewrite it when one status cannot accurately describe both.
Governing boundary
Public
histdata.com-toolsowns the scientific substrate and neutral contracts needed to describe or verify downstream usage. It does not own concrete proprietary:The autonomous economic-release forecasting factory #556 is an upstream public data-producing subsystem and is therefore not reclassified merely because it uses ML internally. This issue concerns downstream learned trading/decision research over the final substrate.
Issue mutation rules
For every reviewed issue:
private_execution_requiredis not confused withpublic not_implemented.Handoff manifest
#721's handoff bundle must expose the public contract versions that downstream private implementations are expected to reference. It must not embed private model artifacts.
Acceptance
Every late ML issue has machine-readable ownership at requirement granularity; #721's dependency graph contains only public responsibilities; private execution work remains traceable for the new repository but cannot prevent the public data/simulation substrate from reaching terminal handoff; no learned-model operational state is silently required to reside in
histdata.com-tools; and #691 proves zero orphaned requirements after the editorial split.