The blog post argues that the term "kilotokens" should be denoted as "kTok" with a lowercase 'k', clarifying the correct SI unit usage. It points out inconsistencies in documentation related to input token pricing for models provided by OpenAI, specifically addressing the misuse of units and highlighting errors in their documentation.