Marven, Aurel

Toward Safe Intelligence: Non-Agentic Architectures as a Structural Alternative to Objective-Based AI Alignment

2025 · OSF Preprints (OSF Preprints)

This preprint introduces non-agentic architectures as a structural alternative to objective-based AI alignment. It argues that systems without internal objectives, temporal projection, or narrative continuity exhibit intrinsically safer computational dynamics than agentic optimizers. The paper …