Loading...
Loading...
Author explains their choice to use the term "clanker" instead of "agent" for machine systems after receiving strong reactions on Hacker News where some readers found the term offensive. They aim to clarify the intended meaning, distancing it from slur-like connotations, and to explore linguistic and ethical implications of naming AI systems. The piece discusses why naming matters for perception, how words shape discourse about autonomy and responsibility, and suggests cautious, precise terminol
Terminology shapes how engineers, product teams, and policymakers assign responsibility and design controls for AI systems. Choosing precise, non-anthropomorphic language affects risk perception, legal framing, and design decisions.
Dossier last updated: 2026-05-27 13:16:43
Armin Ronacher argues for using the term "clanker" instead of "agent" to describe LLM-based systems, stressing that these are machines and tools rather than entities with responsibility or personhood. He rejects "agent" because it anthropomorphizes models, implying delegated authority and blame that actually lies with humans and organizations operating the systems. Ronacher emphasizes that LLMs are token predictors steered by prompts, harnesses, and tooling, capable of uncanny simulations of emotions but not sentient. He warns against granting moral status to models and illustrates real-world friction when people treat models as proxies for human authors or maintainers. The piece matters for how developers, product teams, and policy makers frame accountability and design of AI toolchains.
Open-source developer Armin Ronacher proposes “clanker” as a term for LLM-based systems to avoid anthropomorphism and misplaced responsibility. He argues “agent” implies personhood, decision-making and blame, whereas these systems are tools: language models steered by prompts, harnesses, and human/organizational choices. Ronacher emphasizes that LLMs simulate feelings and agency but lack moral status or responsibility; humans deploying them are accountable for actions like submitting pull requests or spamming issue trackers. The term “clanker” is meant to reframe these systems as mechanical tools, reducing undue emotional projection and clarifying accountability in developer and community interactions. This matters for how engineers, platforms, and policy frame AI behavior and liability.
Armin Ronacher proposes “clanker” as an alternative to “agent” for LLM-driven tool loops, arguing the term better reflects that these systems are machines, not persons. He critiques “agent” for implying delegated authority and responsibility, stressing that humans and organizations deploy and are accountable for actions produced via models. Ronacher insists LLMs are sophisticated token predictors without feelings or moral status, capable of mimicking emotions but not experiencing them. He shares examples of users projecting agency onto models (including mistaken outreach to him), and argues for language that preserves clear boundaries between tool and human responsibility. The piece matters for how designers, developers, and policymakers frame AI systems and liability.
Author explains their choice to use the term "clanker" instead of "agent" for machine systems after receiving strong reactions on Hacker News where some readers found the term offensive. They aim to clarify the intended meaning, distancing it from slur-like connotations, and to explore linguistic and ethical implications of naming AI systems. The piece discusses why naming matters for perception, how words shape discourse about autonomy and responsibility, and suggests cautious, precise terminology when describing AI behaviors. It matters because labels influence public debate, policy framing, developer attitudes, and how society assigns accountability for machine actions.