Gary Marcus Links OpenAI Pricing Idea to Customer Pushback

Gary Marcus posted that OpenAI is considering a pay-when-it-works model for its enterprise AI agents. He stated that the approach, which could require the company to absorb costs on failed outputs, stems from customers recognizing that the systems often produce unusable results and refusing to pay for them. The post references a generated headline about OpenAI testing pay-on-success pricing. Marcus, identified in the packet as a critic of deep learning and LLMs, made the claim on X.