I gave both ChatGPT and Claude exact same prompt and none did anything close to what author was claiming. The blogpost could just be poor AI writing.