Anthropic Faces Scrutiny Amid Claude Code Performance Crisis and User Backlash
Published by AINave Editorial • Reviewed by Ramit
Amid a turbulent period marked by user dissatisfaction, Anthropic has publicly acknowledged a series of engineering missteps that led to a decline in the performance of its AI tool, Claude Code. Following complaints and accusations of gaslighting from its user base, the company’s admission of three distinct technical errors reveals the complexities of rapidly scaling technology against the backdrop of competing interests in the AI industry.
The Root of the Decline
A month-long controversy erupted within Anthropic, one of the leading AI labs with a valuation of approximately $380 billion, as users experienced significant drops in Claude Code's performance. Initially, executives attributed the issues to user feedback and latency improvements. However, detailed investigations led to the identification of three key engineering missteps: a latency-reduction change on March 4, a bug introduced on March 26 that caused the model to forget reasoning history, and an April 16 system prompt that limited responses to 25 words. Collectively, these oversights effectively degraded the quality of output users had come to expect.
As a result of these performance issues, many users reported feeling misled, with some choosing to cancel their subscriptions. This unrest jeopardizes Anthropic’s substantial user loyalty, particularly among developers who have fueled its success and contributed to its annual recurring revenue climbing to approximately $30 billion.
User Trust at Stake
Anthropic's predicament is compounded by increasing competition, particularly from OpenAI, which recently released GPT-5.5. The stakes are high as both companies gear up for potential public stock offerings. Amid this competitive atmosphere, users detailed experiences of confusion and sourcing blame, with frustrations boiling over into public forums.
A representative from Anthropic admitted that demand for Claude Code had surged, stretching the company's infrastructure. Market experts have suggested that failure to secure sufficient computing resources may underscore a strategic misstep on the part of Anthropic, as competitors secure multibillion-dollar data center deals.
Industry Implications
Simultaneously, the ramifications of the performance decline extend beyond mere user dissatisfaction. Detailed coding security analyses revealed that Anthropic's recent output exhibited increased vulnerability rates, with a reported 52% of coding tasks tested by Veracode leading to vulnerabilities. In comparison, OpenAI’s models averaged around 30%. The long-term implications for novice developers utilizing Claude Code could be severe, as flawed outputs could result in serious defects within production code.
Moreover, the ongoing backlash may drive users toward OpenAI's offerings, which currently enjoy widespread popularity among developers; OpenAI has revealed expansive metrics showcasing 4 million active Codex users and 900 million weekly active users on ChatGPT.
Moving Forward
While Anthropic has achieved impressive growth metrics, its path forward requires careful navigation of user trust and resource allocation. Following the recent incident, the company has committed to greater transparency, promising users that the fixes have been made and reiterating the importance of listening to customer feedback. Yet, whether this will be enough to mend the fractured relationship remains uncertain amidst a race for dominance in the AI landscape. In the coming months, Anthropic faces the dual challenge of stabilizing its technology while expanding its infrastructure to meet user demands.