Andrej Karpathy claims X has become so toxic that he now prefers spending time on Anthropic’s Slack; Elon Musk responds

Twitter, also known as X, is facing growing criticism, partly due to its alleged toxicity. Numerous user complaints suggest dissatisfaction with X’s algorithm. The latest person to join this chorus of criticism is Andrej Karpathy, an OpenAI co-founder and former Director of AI at Tesla who recently joined Anthropic. In a recent post, Karpathy noted that X has become more toxic than ever, attributing the platform’s recommendation algorithm with fostering hostile behaviour. Surprisingly, Elon Musk, the owner of X, fully agreed.

In a recent thread on X regarding the “Claude Tag” workflow newly introduced by Anthropic, Karpathy—who joined the company earlier this year—argued that the new system represents a major shift in how people interact with AI. According to him, many users misunderstood his comments and dismissed the idea without reading past the headline.

In his original posts, Karpathy maintained that Claude Tag is not merely another chatbot feature or Slack bot. Instead, he described it as a new way of working with AI, in which the assistant integrates with various tools, memory systems, computing environments, and workflows, making the experience feel more like collaborating with a teammate than chatting with a bot. However, his post quickly sparked a debate on X. While some users agreed that AI is evolving beyond standalone chatbots into workplace assistants that operate across various tools, others mocked the comparison and questioned whether the feature was truly as extraordinary as Karpathy claimed.

In response to the criticism, Karpathy noted that many people had reacted without reading the full explanation. “It’s true; I think quite a few people on the timeline didn’t get past the headline and drew completely wrong conclusions and comparisons,” he wrote, adding that Claude Tag shouldn’t be dismissed as “a mediocre Slack bot.”

The discussion soon shifted from AI to the state of conversation on the X platform itself. In a subsequent post, Karpathy wrote: “I’ve been on Twitter for nearly two decades, so I can say with confidence that it has never been this toxic or resembled Reddit so much. I believe the algorithm actively fosters this situation and that users are being ‘RL’d’ [subjected to reinforcement learning] because of it. That’s the main reason I log in and post less frequently.”

By using the term “RL’d,” Karpathy was referring to the concept of reinforcement learning, suggesting that X’s recommendation algorithm rewards users for posting content that sparks outrage, criticism, or controversy, as such posts generate higher engagement and visibility.

Musk replied directly to the post with a brief comment: “We need a complete algorithm overhaul.”

Musk did not detail the changes he intends to implement in X’s recommendation system, but his comment reaffirmed his long-standing stance that the platform’s algorithms must evolve to improve the quality of conversation. Since acquiring Twitter in 2022 and rebranding it as X, Musk has repeatedly emphasised the need to make the platform’s recommendation systems more transparent and improve the overall user experience.