A months-old but until now overlooked study recently featured in Wired claims to mathematically prove that large language models “are incapable of carrying out computational and agentic tasks beyond a ...