Coding Agents Are Outliers (vivekhaldar.com)

Recent advancements in AI have highlighted the success of coding agents, particularly since the introduction of CLI coding agents like Claude Code in early 2025. These agents excel at complex tasks such as writing, debugging, and deploying code, thanks to characteristics intrinsic to programming: deterministic outputs, clear success metrics, and robust feedback loops. However, the article argues that coding agents are outliers in the broader AI landscape, as the unique factors contributing to their effectiveness are not easily replicated in other specialized domains, such as healthcare or law. The article emphasizes the importance of understanding the distinction between workflows and agents. While coding agents leverage rich domain knowledge and feedback from their creators—many of whom are the programmers building the tools—they excel in a context that is largely absent in other fields where knowledge is often tacit and less documented. Building effective agents in these other domains involves grappling with this tacit knowledge problem, making it challenging to translate the success seen in coding to broader applications. As a result, while coding agents showcase impressive capabilities, they should not set unrealistic expectations for the development of AI agents in other industries.