Foundations

This part establishes the worldview of the Fieldbook before introducing specific architecture or implementation choices.

Chapters

• What Makes a System AI-Native? — defines AI-native software, distinguishes it from agentic systems, and identifies the engineering commitments the design requires.
• Workflows, Agents, and the Autonomy Spectrum — distinguishes control structures and uses a four-part delegation profile to calibrate autonomy.
• Context Engineering: Giving the Model What It Needs — explains how to select, structure, and govern the information available to a model at each inference step.
• Uncertainty and Variable Model Behavior — locates uncertainty, defines acceptable variation, and shows where hard constraints must be enforced around model behavior.
• Evaluation: How Do We Know the System Works? — turns intended behavior into evidence across system outputs, execution, and outcomes.
• Agent Engineering Principles — synthesizes the series into nine durable principles for design and review.

Reading sequence

The sequence moves from identifying an AI-native system, to allocating control, supplying context, reasoning about variable behavior, evaluating results, and applying the combined principles in an engineering review.