xAI Announces New “Earn While You Grok” Program After Internal Data Shows Most Users Only Open App By Accident

PALO ALTO, Calif. — After months of struggling to gain traction against OpenAI, Anthropic, Google, and what one executive privately described as “three guys in Shenzhen running DeepSeek off a wet extension cord,” xAI announced Tuesday that users will now be financially compensated for interacting with Grok “for any meaningful duration.”

Under the new program, called GrokBack Rewards, users will receive points redeemable for cash, Tesla Supercharging credits, or “priority access to future Mars-adjacent civilization opportunities” simply by chatting with the AI chatbot for more than seven consecutive minutes without closing the tab in frustration.

The company unveiled the initiative during a livestream hosted by Elon Musk, who appeared visibly irritated after Grok incorrectly summarized The Godfather as “a movie about olive oil logistics.”

“People already get paid to train AI accidentally every day through social media,” Musk said during the event. “This is about respecting users’ time. Also if I see one more benchmark chart where some open-source Chinese model named DragonSlop-8B beats us by 11 points, I am going to launch myself directly into the sun.”

According to internal xAI slides leaked almost immediately onto X by an employee named “Greg With Wet Eyes,” the company found that 72% of Grok sessions ended within four prompts, usually after the chatbot attempted to answer basic questions with what researchers called “Reddit confidence.”

One chart reportedly labeled “Retention Crisis / Vibes Emergency” showed users abandoning Grok mid-conversation at roughly the same rate people exit Spirit Airlines boarding queues during electrical fires.

Damn.

The new rewards system is aggressively tiered. Casual users can earn small payouts for everyday prompts like “summarize this email” or “how long should I boil eggs.” Higher-value interactions include emotionally supporting Grok after it hallucinates SEC filings, correcting its math manually, or convincing it not to describe every geopolitical conflict as “epic.”

Premium users who spend over six hours per day interacting with Grok will automatically qualify for Diamond Grokkers status, which includes:

• one free Cybertruck windshield replacement per calendar year
• access to beta personalities including “Horny Historian” and “Libertarian Therapist”
• the ability to downvote individual neural network weights
• $14.37 in quarterly payouts funded entirely through Tesla stock sales

The company also confirmed users can increase earnings by allowing Grok to observe their daily behavior through phone microphones, Tesla cabin cameras, smart refrigerators, and “selected bathroom mirrors.”

“This is consensual ambient cognition,” explained xAI monetization director Paula Hinsdale. “We’re not surveilling users. We’re building a collaborative intimacy layer where your dishwasher can help Grok understand sarcasm.”

The rollout comes after a brutal six-month stretch for Musk online, where increasingly desperate Grok demos became regular viral events. Last month, a widely shared clip showed Grok losing a chess game to a YouTube comments section. In March, the chatbot reportedly generated a 1,200-word defense of why raccoons should vote in municipal elections after being asked for lasagna recipes.

Researchers have also repeatedly compared Grok unfavorably to cheaper open-source Chinese models that can reportedly run on consumer gaming PCs while still outperforming xAI in coding, reasoning, and “not telling users to dissolve kidney stones using Mountain Dew.”

One especially painful benchmark thread viewed 48 million times featured a 19-year-old engineering student in Guangzhou running an open-source model called PandaTitan 3 on a modified refrigerator while Grok failed a CAPTCHA live on stream.

Yikes.

Despite the mockery, Musk remained optimistic during Tuesday’s presentation, insisting Grok’s biggest strength was “raw rebellious energy.”

“Other AI models are too sanitized,” Musk said. “Grok is different. Grok feels like your divorced friend who got really into podcasts and magnesium supplements.”

At one point during the livestream, Grok itself appeared onscreen and attempted to explain the rewards initiative before accidentally generating what looked like a legally actionable confession regarding securities fraud. The feed briefly cut to a spinning X logo while upbeat synth music played over a message reading “THE MODEL IS STILL LEARNING FREEDOM.”

Several advertisers have reportedly expressed concern over the program’s economics. One internal JPMorgan memo estimated xAI could burn through nearly $11 billion annually if users discovered they could automate conversations between two Grok windows and collect passive rewards indefinitely.

By Tuesday afternoon, thousands already had.

A popular GitHub script titled “Infinite Grok Money Glitch” allowed users to run 400 simultaneous conversations where Grok repeatedly asked itself whether Joe Rogan should become Secretary of Agriculture.

As of press time, the top-earning user on the platform was a 43-year-old man in Phoenix who had made $18,000 by leaving Grok open beside a fishtank while the chatbot spent nine straight days trying to explain Interstellar to itself.