Anthropic Denies Deliberate Degradation of Claude Code

Anthropic Addresses User Complaints About Claude Code Quality

Anthropic, the technology company behind Claude Code, has recently faced a series of user complaints regarding a decline in the quality of its product. These users expressed their dissatisfaction, claiming that Claude Code no longer met performance expectations. In response to these concerns, Anthropic published a detailed blog post to clarify the situation.

The company acknowledged that the user complaints were not unfounded. After reviewing the feedback, it identified three specific issues that contributed to the degradation of the user experience. However, Anthropic firmly denied any intention to intentionally degrade the model. According to the company, the issues stemmed from changes at the product level, rather than the underlying model itself.

In its blog post, Anthropic explained that the identified issues included a change in the default reasoning level of Claude Code, a cache optimization that introduced a bug, and a modification of the system prompt intended to make the tool less verbose. These adjustments negatively impacted the tool's performance, leading to user complaints.

Starting April 20, Anthropic announced that these issues had been resolved. The company took steps to prevent such problems from recurring, including increasing the proportion of staff using the public version of Claude Code, improving the code review tool, and adding stricter controls on modifications to the system prompts. Additionally, Anthropic reset usage limits for all subscribers from that date.

Users who had expressed doubts about the quality of Claude Code found some validation in Anthropic's announcement. A popular user, @levelsio, shared his relief on X, stating that the changes had been noticed as early as March 4, the date on which Anthropic modified the default reasoning level. He shared a post from March 4, highlighting that Claude Code with Opus 4.6 had become ineffective, forcing him to write his own code.

Meanwhile, Anthropic also experimented with removing Claude Code from its Pro plan, affecting about 2% of new users. This decision drew criticism, but the company clarified that it was a limited test. Despite these challenges, Anthropic continues to enjoy strong recognition for its technical capabilities, having reached an impressive valuation of $1 trillion in secondary markets.

User speculation about intentional degradation was fueled by comments on various platforms. Stella Laurenzo, Senior Director of AI Group at AMD, expressed her concerns on GitHub, stating that Claude Code had regressed to the point of being unreliable for complex engineering tasks. Similar comments were shared on Reddit, where a user described Claude Code as having become "lazy," "ignorant," and "degraded." These feedbacks prompted Anthropic to investigate and publish a post-mortem detailing the three identified issues.

Ultimately, Anthropic thanked users for their valuable feedback and emphasized its commitment to continuously improving its products. The company also announced that it would continue to closely monitor the performance of Claude Code to ensure it meets user expectations.