The author is missing the whole point. The exercise answers the implicit question: “can Ai do real work in this world? Can it replace workers, can it run a busi

The author is missing the whole point. The exercise answers the implicit question: “can Ai do real work in this world? Can it replace workers, can it run a business?” And the experiment answers that question decisively “no it cannot”. Why not? Read the article and ponder the seven to eleven distinct ways the AI goes wrong. Every AI enabled workflow will fail the same way. It even raises the question “is this technology even deserving of the title AI at all?” (If you ask me, probably not) It lacks common sense, it lacks a theory of the world it lacks the ability to identify errors and self-correct. It lacks the ability to introspect. Now consider the thousands of people out there pushing the idea of fully autonomous AI “agents” doing work in the real world. Are we really for that? Again the exercise answers the question definitely. “No, the technology is not ready and cannot be trusted, and maybe in the current incarnation (Namely LLM based) cannot ever be trusted” Now consider the tens of millions of people who think the technology IS ready for that. Anthropic is in the business of studying AI safety, publishing research, examining what AI is and how far we can trust it. They did that job smashingly.