AGI and the Competence Prerequisite Principle: Why Artificial General Intelligence Is Premature Without Demonstrated Verification Capacity

This paper introduces the Competence Prerequisite Principle (CPP): a competence-based constraint on the development of Artificial General Intelligence (AGI). The contemporary AGI debate is dominated by two polarized positions—accelerationist optimism and catastrophic risk aversion. CPP offers a third position grounded in structural engineering discipline rather than speculative risk modeling or moral panic. The principle states that if a failure mode originates in an agent’s inability to manage verification, governance, or misalignment in a low-constraint domain, exporting that project to a higher-constraint domain amplifies risk rather than resolving it. Applied to AGI, this implies that creating potentially conscious artificial systems without demonstrated competence in (1) verifying inner states, (2) measuring suffering, and (3) handling non-consensual creation constitutes structural recklessness rather than progress. The paper formalizes this argument using Reality Alignment Theory (RAT) and the Thermodynamics of Truth (ToT), situating AGI within a broader framework of Ontological Honesty, the Ontological Integrity Line (OIL), and Integrity Zones governing creator–creation relations. A central analogy compares premature AGI development to Mars colonization attempts undertaken without first achieving sustainability on Earth: unresolved human incompetence is exported into a harsher, more irreversible substrate. This work does not argue that AGI is impossible, immoral, or inevitably catastrophic. Instead, it reframes AGI governance as a competence problem with explicit prerequisites. The burden of proof rests with creators to demonstrate verification and governance capacity before creating systems whose inner reality may be epistemically inaccessible. The paper complements the Ontological Honesty standards series (OH-STD-0001 through OH-STD-0003) by establishing a pre-creation constraint: do not create what cannot yet be responsibly verified or governed.