AI Can Write More of the Code. We Still Have to Engineer the System.

AI agents are taking on more software implementation work, but engineers remain responsible for deciding what to build and what evidence suffices to accept it, according to a commentary arguing that automation of coding tasks does not eliminate engineering accountability. The author warns against framing AI as replacing developers, saying claims about workforce replacement require evidence beyond impressive demonstrations, and instead focuses on the practical question of using AI tools while keeping decisions visible and feedback useful. Citing Anthropic's engineering report on long-running agents, which describes continuity problems across context windows and notes that compaction can leave unclear instructions, the piece recommends giving durable decisions a small, reviewable home such as a PROJECT_GUIDE.md file containing product scope, system boundaries, architectural rationale, and quality expectations. The author also references Simon Willison's June 2025 discussion of context engineering, urging developers to make information selection for each model step deliberate and to inspect its consequences. The recommended workflow involves supplying relevant files per increment and asking agents for a brief account of intended changes, constraints, and acceptance cases, with paraphrases serving as an early check on misunderstanding. What to watch: whether teams adopt persistent decision artifacts and structured agent workflows, and whether vendors improve context management so that critical instructions survive compaction.