Agentic Autonomy Levels

This article discusses the evolution of agentic engineering from simple prompting to operating autonomous agents. It introduces a two-dimensional framework for measuring autonomy: agency (how independently a single agent acts) and orchestration (how multiple agents are coordinated). The author critiques single-axis ladders like Steve Yegge's and argues that modern multi-agent setups require separate consideration of these dimensions. The article covers low-to-high agency levels (from suggesting actions to goal-driven experimentation) and low-to-high orchestration (from single agent to manager agents delegating to subagents). It is highly relevant to IT/technology, focusing on AI agents, software engineering, and autonomous systems.