Over the next 5–6 years, the underlying capabilities that will retain their value for AI engineers include: understanding retrieval and information architecture, evaluation and experimentation ability, system design and engineering, and the judgment to assess the boundaries of model capabilities.