The article provides no solid evidence that "AI is working" for the author.

At the end of the day this article is nothing but another piece of conjecture on hackernews.

Actually assessing the usefulness of AI would require measurements and controls. Nothing has been proven or disproven here