A public exchange between former Tesla AI director Andrej Karpathy and X owner Elon Musk has reignited debate over whether the platform’s recommendation algorithm is rewarding outrage over meaningful conversations, with Musk saying the algorithm needs a complete overhaul.
The discussion began after Karpathy, one of the most respected voices in artificial intelligence, shared his frustration with how X has evolved.
“I’ve been on Twitter for almost 2 decades now so I can say with confidence that it has never been this toxic and Reddit-like. I think the algorithm actively encourages it and people get RL’d by it. Main reason I’ve been coming here less and posting less.”
Karpathy suggested that the platform’s recommendation system is amplifying confrontational and polarizing content, making the experience less enjoyable even for longtime users.
Just hours later, Elon Musk responded with a brief but notable reply:
“We need a complete overhaul of the algorithm.”
The exchange is significant because it comes from two people who have worked together closely at Tesla and who have deep expertise in AI systems.
A long-running criticism of social media
Karpathy’s comments echo concerns researchers and users have raised for years: recommendation algorithms often prioritize posts that generate strong emotional reactions because those posts drive engagement. While this can increase time spent on a platform, critics argue it also rewards outrage, conflict, and divisive content.
Many users have claimed their “For You” feeds have become increasingly argumentative, with controversial posts frequently receiving outsized visibility.
Musk has acknowledged the problem before
This isn’t the first time Musk has criticized X’s recommendation system.
Last year, he admitted the algorithm often mistakes outrage for interest. For example, if someone shares a post because they disagree with it, the system may incorrectly assume they want to see more similar content.
Earlier this year, Musk went even further, calling the algorithm “dumb” and saying it requires “massive improvements.” He also pledged greater transparency by publishing updates to the recommendation system as X continues rebuilding it around newer AI models.
Musk has repeatedly said his goal is to maximize what he calls “unregretted user-seconds”—time users enjoy spending on the platform, rather than engagement driven by negativity alone.
Will X actually change?
Musk’s latest response suggests the company recognizes that the current recommendation system still isn’t delivering the experience it wants.
Whether a “complete overhaul” results in less toxic conversations remains to be seen. Balancing free expression with healthier recommendations has become one of the biggest technical challenges facing modern social media platforms.
For now, Karpathy’s criticism—and Musk’s agreement—highlights that even those closest to the technology believe X’s algorithm remains a work in progress.
Add Comment