Skip to main content

Advanced Roadmap (Horizon 2/3)

This page isolates advanced architecture concepts that may appear in research notes or older documents. They are intentionally separated from the operator docs because they are not guaranteed to be active runtime enforcement in the current main branch serving path.

Status tags used here:

  • [Status: Research] - explored, prototypes may exist, not enforced by default
  • [Status: Planned] - design intent only

Concrete-ML / FHE / zkPC proofs​

[Status: Research]

Goal: cryptographically verifiable enforcement of F9 Anti-Hantu (and related floors) without exposing raw prompts.

Reality today:

  • Mainline docs treat zkPC as a governance concept and (where present) an audit receipt model.
  • Do not assume Fully Homomorphic Encryption circuits or zero-knowledge compliance proofs are enforced in production by default.

Postgres.ai Database Lab Engine (DLE) thin-clone rollbacks​

[Status: Planned]

Goal: zero-cost rollback sandboxes to strengthen F1 Amanah for high-stakes database operations.

Reality today:

  • Postgres is supported and recommended for durable VAULT999 persistence.
  • Do not assume ZFS thin clones or DLE orchestration is integrated into the default pipeline.

AutoGen Trinity consensus​

[Status: Research]

Goal: multi-agent deliberation loops for capability exploration, with APEX veto preserved.

Reality today:

  • arifOS architecture documents reference multi-engine separation (AGI/ASI/APEX) and tri-witness checks.
  • Do not treat AutoGen as a required runtime dependency or an always-on enforcement component.

Prefect ControlFlow / long-running workflows​

[Status: Planned]

Goal: deterministic orchestration and human-in-the-loop pausing for 888_HOLD and long-running governance flows.

Reality today:

  • 888_HOLD is a documented governance state.
  • Do not assume ControlFlow orchestration is present unless your deployment explicitly includes it.