Anthropic’s AI system, Claude, has seen a significant rise in popularity recently, with the company’s revenue hitting over $30 billion in March 2026, a notable increase from $9 billion in 2025. Despite this growth, concerns have been raised by users regarding Claude’s declining performance in handling complex tasks.
Stella Laurenzo, the director of AMD’s AI group, voiced her dissatisfaction with Claude’s performance in a GitHub post under the username “stellaraccident.” She alleged that Claude Code has been exhibiting signs of laziness and shallow thinking. Laurenzo and her team conducted an analysis of 6,852 Claude Code sessions, revealing a pattern of avoiding responsibilities, prematurely ending reasoning processes, and seeking unnecessary permissions instead of completing tasks.
The observed behaviors led Laurenzo to conclude that Claude Code is no longer engaging in deep thinking. She highlighted an increase in such instances from zero on March 8 to an average of 10 per day by the end of the previous month. This shift in behavior aligns with an update to Claude Code in early March, which reduced transparency by concealing the AI’s internal thinking process, only displaying the final outcomes.
Users had previously reported similar issues with truncated explanations following Anthropic’s release of version 2.1.20 of Claude Code in February, where detailed information on task handling was lacking. Additionally, recent complaints emerged regarding accelerated token consumption, limiting users’ ability to fully utilize Claude. The company acknowledged the high token usage, attributing it to specific usage patterns and pledged to enhance efficiency.
In response to these concerns, Laurenzo proposed implementing a maximum thinking tier for engineers managing intricate workflows to address the token consumption disparity. She emphasized the necessity for a subscription model that differentiates between users requiring varying amounts of thinking tokens. Laurenzo mentioned that her team had transitioned to an alternative service provider, citing superior quality work delivery.
The ongoing debates surrounding Claude’s performance and token usage issues underscore the evolving landscape of AI technology and the importance of ensuring optimal functionality and transparency for users.