Clanker: A Word For The Machine

Armin Ronacher defends his use of the word 'clanker' as a deliberate alternative to 'agent' for LLM-based tool loops. His core argument is that anthropomorphizing AI systems is harmful because it diffuses human responsibility — if a coding assistant drops a database or spams an issue tracker, the human who deployed it is responsible, not the machine. He argues that LLMs are sophisticated token predictors, not sentient beings, and that treating them as a protected class or extending moral consideration to them risks devaluing actual human suffering. He also acknowledges a problematic use of 'clanker' in online communities where it is used to replay racist imagery through a sci-fi lens, which he explicitly rejects. The post calls for language that maintains a clear boundary: humans carry responsibility, machines are tools.