Agent Native Architecture

This is a philosophical framework for building applications where agents accomplish tasks instead of executing predefined code paths. The core idea is "parity": if a user can do it through your UI, the agent should be able to do it through tools. It's built on the observation that Claude Code's bash-and-files-in-a-loop architecture works for way more than just coding. You'd reference this when designing agent-based features or rethinking how your app exposes capabilities to LLMs. The principles are sound but fairly abstract. Think of it as architectural guidance rather than implementation code. Useful if you're in the "how should this actually work" phase of building agent-driven software.