kadubon/collective-capability-runtime: CCR 1.8.0: verified reuse and verification-aware growth allocation
Abstract
CCR 1.8.0 adds an explicit verified_growth_v1 policy while preserving the legacy v1 optimizer, signatures, admission, approval and lease boundaries. Replayable service/cost accounting, delayed reuse, immutable assets, receiver transfer qualification, and lifecycle invalidation/revalidation. Exact bounded dependency-bundle planning, independently checked reservations, verifier-service investment, conservative shared quotas and frozen restricted comparisons. Pinned, evidence-only CAIT/VEK/ALT interchange and packaged offline scenarios using ephemeral keys and real task/workcell transitions. Install with python -m pip install 'collective-capability-runtime[optimizer]==1.8.0', then run ccr optimizer growth-example --json. Distributed dependencies remain optional; Python 3.10+ support is retained. The supported initial checkpoint is empty evidence. Finite forecasts are not observations, asset count is not capital, and joint capacity and causal/statistical improvement remain unresolved. The reoptimized permitted reuse alternative matches the candidate in the packaged null comparison. No companion repository or production provider is modified or contacted. Validation and exact build/public-index receipts: https://github.com/kadubon/collective-capability-runtime/blob/main/docs/verified-growth-validation.md . Distribution assets and SHA256SUMS must match the same Trusted Publishing build; public-index verification is a separate post-publication state. No external empirical collective-intelligence acceleration experiment was performed for this release. The implementation supports evidence-bound coordination, bounded planning, qualified reuse, and reproducible software-level testing. It does not establish causal endogenous capability reproduction, AGI/ASI, indefinite growth, or correctness outside the declared evidence and model boundaries.
