
Anthropic has acknowledged user concerns regarding the performance of Claude Code, its AI product, confirming significant quality issues. In a blog post released on Thursday, the company identified three primary factors contributing to the decline in user experience.
Key Findings on Claude Code's Performance
Anthropic's investigation revealed that the underlying model of Claude Code remained intact; however, several product-level modifications had negatively impacted its performance. The company emphasized that any degradation was unintentional.
Major Issues Identified
Change in Claude Code's default thinking level.
Bug introduced by a cache-optimization adjustment.
System prompt adjustments aimed at reducing verbosity.
As of April 20, Anthropic has addressed these issues and implemented measures to prevent future occurrences. The company has begun resetting usage limits for all subscribers, thanking users for their feedback on the platform.
User Reactions and Speculation
Many users expressed frustration over perceived declines in Claude Code's reliability. Users on GitHub and Reddit commented on the tool's recent performance, with some claiming it had become "lazy" and less competent in handling complex tasks.
Stella Laurenzo, a senior director at AMD, stated that Claude had regressed significantly. Another user, reporting similar sentiments, noted a marked decline since a specific change was made on March 4.
An Overview of User Experiences
Users reported that Claude Code became less effective at performing intricate engineering tasks.
Some felt validated by Anthropic's admission of the issues.
Despite the recent setbacks, Anthropic continues to receive praise for its technological achievements and has recently achieved a valuation of $1 trillion in secondary markets. The company also faced scrutiny concerning a trial removal of Claude Code from its Pro plan, which affected about 2% of new users.
Anthropic's commitment to rectifying these issues indicates its dedication to maintaining the quality of Claude Code, assuring users of improved functionality in the future.