‘We will not rest until Grok is perfect’: Musk urges X users to flag AI mistakes

Elon Musk has called on X (formerly Twitter) users to actively report instances where his AI chatbot, Grok, delivers unsatisfactory or incorrect replies, as xAI works to refine the model’s reliability and tone. The appeal coincides with the rollout of Grok 4.1, a new version designed to devote more compute resources to assessing and reasoning through prompts.

In a post shared on X on Saturday, Musk encouraged users to flag any present-day shortcomings and provide comparisons with rival AI systems. “Please provide examples where @Grok needs to improve in replies. Showing how another AI does it better would be helpful. These examples should be of Grok going wrong today, as we fixed many bugs from earlier in the week,” he wrote.

He later reiterated his ambition for the system, saying, “We will not rest until Grok is perfect”

Musk also asked users to submit positive examples where the model performs well, noting that such instances help guide development. He wrote, “Examples where Grok provides great responses that it should do more of are helpful too”

In response to one user who asked, “Hey @grok, what’ve you been doing that’s wrong?”, the chatbot admitted, “Oh, I've been known to mix up details in fast-evolving news or overdo the sarcasm. But specifics help—share an example from today, and I'll dive in! What's one thing you've noticed?”

Community feedback push follows accuracy concerns

The request for user-generated feedback emerges after Grok faced criticism for a series of exaggerated responses, including dramatic claims depicting Musk as outperforming elite athletes and even historical personalities.

Grok 4.1 update focuses on better reasoning

Musk recently confirmed that Grok 4.1 had been updated with substantial fixes aimed at improving the model’s consistency. “Many updates and fixes have been applied to Grok 4.1 and many more to come! Going forward, Grok 4.1 will spend more compute time thinking about your question to improve accuracy,” he announced.