Exodus and Congestion
The report describes a user migration: for various reasons (word of mouth, features, disappointment with a certain provider), a batch of ChatGPT users poured into Claude, and the freshly arrived load put pressure on Anthropic's capacity and stability, so existing users experienced tightened limits and slower responses. This is a textbook growing pain: users voting with their feet proved the product's appeal, but the sudden traffic also exposed the elastic limits of the infrastructure. Winning users and serving users well are two different exams.
The Low-Loyalty Era of AI Users
Buried underneath this story is a reality that holds for every lab: user stickiness in AI products is lower than you'd think, switching costs are near zero, and a single price hike, a single reputation collapse, or one more usable competitor can trigger a mass migration. This is a double-edged sword for vendors—today's winner may be tomorrow's loser, and nobody dares sleep on their user base. For users it's good news: ample mobility means continuous bargaining power, and whenever a service gets worse or pricier, you just switch. For Anthropic, this congestion counts as a happy stress test—but if it can't catch the growth, the reputation it loses will come faster than the users it gains.
via: Hacker News