Comment by uriegas

I was thinking the same thing. Maybe is that at the end the author seems to imply that agentic AI will work simply because models have become better regardless of the way we make them agentic (i.e. MCPs, skills, etc).