Anthropic is currently undertaking A/B testing for the subscription quotas of Claude, a scenario in which users subscribed to the same plan might encounter different usage limits. Certain accounts have observed an approximate 20% increase in their usage quotas over a five-hour span. However, the API equivalent value of their monthly available resources has concurrently decreased by around 20%. Recently, the quotas for Claude have been subject to multiple modifications. When combined with the diverse array of models, varying inference intensities, and multiple sets of time window restriction rules, the user experience has exhibited significant fluctuations. Anthropic's strategy is geared towards enhancing the profit margins of its subscription business. Currently, subscriptions account for merely about 10% of its revenue, while consuming over 40% of its inference computing power. In contrast to OpenAI's approach of directly slashing quotas, Anthropic has opted to redistribute computing power across different time windows. This has led to pronounced differences in user experience for individuals with varying usage patterns. Since this test was not communicated to users beforehand and the quota rules governing the subscription service remain opaque, an information asymmetry has emerged. This has sparked concerns among users regarding the fairness of their rights. Users maintain that after making a payment, they should be clearly informed about the stable amount of usage they are entitled to.