Your LLM Is Not a Colleague

The article critiques the growing tendency to refer to LLMs as teammates or colleagues, arguing this language enables blame-shifting when things go wrong. The author, who merges ~650 AI-assisted PRs yearly, distinguishes harmless anthropomorphism from harmful deflection. They highlight how marketing pushes personification (e.g., Devin, Coworker) to boost retention, but this creates a dangerous side effect: responsibility can be transferred to an unaccountable tool. Emphasizing the 'calculator test', the author insists developers must take ownership of AI outputs, as the tool is not a colleague but a tool. The piece offers a sharp perspective on accountability in AI-assisted development.