What PBE is
An execution-governance layer between a proposed action and the system that can make that action real.
Public-safe architecture
Governed execution without blurred authority.
Nemo Flow Systems separates research foundation, execution decision, enforcement, continuity, and proof. The public surface explains the role of each layer without exposing protected mechanisms.
System model
The boundary sits between meaning and consequence.
1. Proposed action
An AI model, agent, or operator proposes a specific bounded consequence.
Proposal only
2. PBE decision
Authority, policy, payload, and relevant state are evaluated before execution.
ALLOW / HOLD / REFUSE
3. Adapter enforcement
Only ALLOW may reach the approved adapter path. HOLD and REFUSE remain blocked.
Bounded consequence
4. Receipt evidence
Decision and outcome are recorded so coherence can be checked independently.
Verifiable record
5. UCA continuity
Mission, checkpoint, evidence, hard stops, and next eligible action remain intact across handoffs.
No authority inflation
Policy-Bound Execution
PBE governs the transition into action.
PBE is a refusal-first execution boundary. It is model-agnostic and separate from the AI that proposes the action.
PBE decides
The decision is explicit: ALLOW, HOLD, or REFUSE.
Adapter enforces
Only the approved consequence path is available after ALLOW.
Receipt chain proves
Decision and consequence are preserved for independent verification.
What PBE is not
It is not an AI model, a prompt filter, a general log system, or a claim that all organizational policy is solved.
Current status
A functional local gate with one controlled real ALLOW use and independently verified receipt evidence.
Continuity without permission
- Preserves exact mission and checkpoint identity
- Preserves evidence and hard stops
- Preserves the next eligible action
- Prevents silent authority expansion across handoffs
- Does not approve execution
- Does not replace PBE
- Does not turn memory into authority
Universal Continuity Architecture
UCA preserves the work without inventing the next step.
UCA is the continuity architecture for governed systems. It keeps mission, proof, state, and hard stops coherent when a session, model, environment, or operator changes.
The continuity foundation completed a private, synchronized checkpoint and successfully supported cold-start resumption of PBE work without drift.
Research foundation
DPI established the governing principles.
Doctrine-Bound Purposeful Intelligence is the foundational research architecture behind the later execution and continuity systems.
Purpose
Action must remain connected to an established mission rather than unconstrained capability.
Authority
Permission must be explicit, bounded, and distinct from the actor seeking to execute.
Evidence
Governance must leave proof that can survive the system that produced it.
Positioning: DPI is the research foundation. PBE is the active execution-governance system. UCA is the continuity architecture.
Use cases under evaluation
Start narrow. Prove the boundary. Expand only with evidence.
Controlled file release
Bounded movement of an approved artifact through a declared adapter path.
Agent tool action
Authority checks before an AI agent invokes a consequential tool.
Infrastructure change
Gated execution before a system modifies production-like state.
Robotic motion
Future bounded motion-control adapters where refusal must remain enforceable.
Only the controlled local file-copy boundary has been demonstrated as a real PBE use. Other use cases are evaluation directions, not deployment claims.
Evidence before expansion
Open the proof ledger