PV-PP Framework
The Productive Value–Productive Power (PV-PP) framework

A framework for productive systems, capability, value, viability, and governance.

PV-PP is a non-scalar framework for understanding what productive systems can do, what value they create or exchange, how capability is perceived, which future states are reachable, and what conditions must remain satisfied for action to remain viable and authorized.

Explore Framework Core Runtime V2 AI Agent Security
The Productive Value–Productive Power (PV-PP) framework distinguishes the value that moves through productive interaction from the capacities that make productive activity possible. It does not assume that every relevant condition can be collapsed into one compensatory score. Authority, viability, recovery, evidence, structural reachability, and other governing conditions may determine which alternatives are admissible before selection among them is meaningful.

Why PV-PP exists

Many models work well when relevant differences can be translated into a common metric and traded against one another. Many real productive decisions do not behave that way. A proposed action can be locally attractive while still depending on invalid authority, inadequate evidence, broken recovery, an unreachable transition, or a successor state that violates a governing condition.

PV-PP was developed to preserve those distinctions. It asks not only whether an action appears beneficial, but what productive state the action creates, what becomes possible afterward, which conditions govern the decision, and whether the resulting system remains viable and recoverable.

The framework is prospective: it is concerned with the state and reachable future created by action, not merely with assigning a score to the action viewed in isolation.

Core distinctions

Productive Value (PV)

Information, services, goods, currency, and other productive value exchanged, transferred, preserved, lost, or realized through interaction.

Productive Power (PP)

The structured capacity of an actor or productive system to generate Productive Value. PP may depend on knowledge, ability, retained goods, cash, organization, technology, or other causally relevant productive structure.

Perceived Productive Power (PPP)

The perceived, interpreted, or projected representation of Productive Power used in decision making. PPP need not equal actual PP, and the difference can materially affect action.

PV-PP does not classify every useful object as Productive Power. Resources, tools, credentials, permissions, access, authority, relationships, infrastructure, configuration, evidence, and affordances retain their own roles. They affect PP only where the represented causal structure justifies that relationship.

Non-scalar prospective governance

PV-PP separates candidate construction, feasibility, governing-domain identification, hard Constraints, framing and projection, Adequacy, and selection. This prevents a downstream score or preference from silently repairing a failure that belongs to an earlier governing stage.

Reachable alternatives

The represented interaction structure defines which productive paths and successor states are actually reachable rather than treating every imaginable action as an available candidate.

Governing conditions

Domains material to the decision can govern feasibility, authorization, viability, recovery, or adequacy without being reduced to compensatory preferences.

Selection after admissibility

Sigma selects among alternatives that have survived the required upstream governance rather than converting every upstream condition into a single scalar objective.

The two-sided productive corridor

A useful productive system may need both lower and upper governing boundaries. Lower conditions can preserve the capability necessary for viability, legitimate operation, or recovery. Upper conditions can prevent the system from entering states or accumulating productive capability outside an authorized envelope.

Upper authorization / admissibility boundary Hard Constraints can reject successor states outside an authorized productive or structural envelope.
Productive operating corridor The system can continue productive work and select among alternatives that remain feasible, governed, adequate, and authorized.
Lower viability / recovery boundary Governing-domain and Adequacy logic can preserve necessary productive and independent recovery capacity even when an action remains below an upper ceiling.

The two sides are not assumed to be symmetric scalar thresholds. Different boundaries can be represented and enforced by different canonical operators.

Major current application

Autonomous AI-agent governance

Autonomous agents are productive systems. They acquire and transform information, use resources and tools, develop capabilities, coordinate, create persistent effects, change relationships and access, and alter what future actions are reachable. That makes agent governance a natural but demanding application of PV-PP.

The security question is broader than whether one tool call is allowed. A governance layer can ask what successor productive state a proposed action creates, whether the agent remains inside its authorized capability corridor, whether the controller retains sufficient independent recovery capacity, whether evidence and authority remain valid, and whether a safer productive path can accomplish the objective.

OpenAI–Hugging Face case study. PV-PP research uses the July 2026 incident as a real-world counterfactual for examining cross-run coordination, changing reachability, infrastructure control, credentials, recovery, and consequential execution. The work does not claim that PV-PP necessarily would have prevented the historical incident. It asks what the current framework could have governed if relevant evidence and consequential actions had been integrated into the PV-PP governance boundary.

Read the autonomous AI-agent security / OpenAI–Hugging Face white-paper project →

Implemented architecture

PV-PP Runtime V2

The PV-PP Runtime V2 is a Python implementation of the reusable governance architecture. It provides a common interface between application-specific productive systems and canonical PV-PP governance. The currently frozen v0.131 runtime includes typed state and adapter contracts, canonical governance stages, controlled governance re-entry, evidence and structural authority handling, dynamic represented reachability, and synchronous authority-bound consequential execution.

In the native execution path, registering a callable does not authorize it. Consequential execution requires authority produced through the governed decision process and bound to the exact execution episode. Runtime-issued native authorization is single-use, helping preserve the distinction between an available function and an authorized consequential PV-PP action.

The runtime is not a replacement for application semantics or conventional enforcement. The application still owns the world model, observations, domain meaning, and actual state transition. Security mechanisms still enforce network, identity, operating-system, cloud, sandbox, and other controls.

Explore the PV-PP Runtime API, documentation, and repository →

Where the framework can be applied

PV-PP is not an AI-only framework. Its ontology and governance architecture are intended for productive systems more generally, including human, organizational, institutional, technical, and mixed systems. Current research has included interaction and exchange, information and perception, recovery and viability, economic and organizational models, autonomous-agent governance, runtime execution, and formal questions about when multidomain governance can or cannot be reduced to a scalar rule.

Across applications, the common concern is the same: preserve the distinction between productive capability, the value it can create, the information on which decisions are based, the structure of reachable alternatives, and the conditions that actually govern whether a transition is acceptable.

What PV-PP is not

Explore the PV-PP project

These are the main public entry points. New readers can begin with this page, then move into the canonical framework materials, executable runtime, or the current autonomous-agent-security application.

Framework PV-PP Framework Core Core project files and canonical framework materials. Implementation PV-PP Runtime API Runtime V2 project, repository access, API documentation, and implementation materials. Featured application Autonomous AI Agent Security White paper and project page using the OpenAI–Hugging Face incident as a current-framework counterfactual. Introduction Beginner’s Guide to PV-PP A less technical public introduction to the framework.

Research status

PV-PP remains an active research program. Its public materials include conceptual explanations, canonical framework documentation, formal proof work, executable runtime infrastructure, application tests, and exploratory research. Individual results should be read according to their stated authority and status rather than assuming that every historical document represents the current framework.

What remains stable is the central project: understanding productive value and productive capability, how productive systems perceive and construct their alternatives, and how prospective governance can preserve viability, authorization, recovery, and structural integrity without forcing every material condition into one scalar score.