arc-policy-runtime-submission

Policy Runtime Submission Package

This folder is a cleaned submission-oriented package derived from a locked working runtime reconstruction.

Design goals:

• keep the behaviorally important runtime logic
• remove snapshots, recordings, archaeology, and diagnostics clutter
• replace old internal branding with generic submission-facing names
• keep only the governance-state pieces actually needed by the bridge

Included Components

• agent_runner.py
  • standalone runner for local verification
• observation_adapter.py
  • converts raw ARC observations into structured runtime state
• policy_bridge.py
  • goal/evidence bridge that scores actions and updates guidance
• runtime_core/
  • the full active ARC runtime package used by the working reconstruction
• governance_state/
  • reduced replacement for the original broader governance runtime package
  • only includes the pieces needed by policy_bridge.py

What Was Intentionally Excluded

• all *.snapshot_*, *.bak, and reconstruction artifacts
• recordings and replay logs
• sidecars, analyzers, and trace-diff tools
• debug JSONL outputs
• historical notes and bug tracker documents
• the unused majority of the original internal governance runtime package

Used Governance-State Pieces

policy_bridge.py only needs these public symbols:

• MemoryCell
• EpistemicStatus
• Goal
• GoalPolicyClass
• GoalStatus
• PrecedenceClass
• SourceType
• compute_goal_weights
• create_initial_state
• create_runtime_support

Those are implemented here via:

• governance_state/cells.py
• governance_state/goals.py
• governance_state/state.py
• governance_state/bootstrap.py
• plus small transitive support files:
  • actions.py
  • branches.py
  • contradictions.py
  • persistence.py

Notes

• This is not yet a final Kaggle notebook package.
• It is the code-shaped core we can submit from.
• The next step is to wrap this logic in a notebook entrypoint that writes submission.parquet under Kaggle competition constraints.