Build AI agents using prompt-native architecture where features are defined in prompts, not code. Use when creating autonomous agents, designing MCP servers, implementing self-modifying systems, or adopting the "trust the agent's intelligence" philosophy.