The Hard Part of AI Coding Agents Isn't Writing Code — It's Preserving Engineering Context

AI coding agents demonstrate proficiency in reading codebases and implementing changes. However, a significant challenge emerges in longer-lived projects: the loss of engineering reasoning between sessions. While the code persists, the underlying justification for validation rules, architectural decisions, or specific requirements often vanishes. This issue is not related to context length, but rather the inability of new agents to understand why certain implementations exist or if they were based on approved requirements or previous assumptions.