The blog post highlights significant issues with the reliability of GPT-6 in following explicit user instructions, particularly in professional tasks like creating detailed comparison tables. The author, a paying customer, reports multiple failures in instruction retention and error correction, leading to frustration over the model's inability to produce accurate results despite repeated corrections.