xAI was founded in 2023, and it’s already changing things radically. Now a key part of a quirky conglomerate that may soon be one of the largest publicly traded companies in the world, it seems like XAI in particular is becoming a distillation of Elon Musk’s mindset as a company. It’s probably not unrelated that all of its non-Musk creators have now left the picture.
Apart from Elon Musk, none of the 12 original co-founders of xAI work for the company as of this week, according to Business Insider. The departure of Ross Nordeen, Tesla’s former manager in the “Autopilot” (driver assistance) division, reportedly makes Musk the last founder.
Two weeks ago, Elon Musk wrote on The departure of nearly half the co-founders was reported within days of XAI’s surprise acquisition by SpaceX in February. The constant drumbeat of the founder’s departure continued until Nordeen’s departure last week.
One might be mistaken in thinking that the content of Grok’s output is the cause of this change.
After all, it was founded in March of 2023, a moment in the culture wars when right-wing enthusiasm over the concept of wokeness was at its peak, before the frenzy died down a bit, as reflected in Google Trends searches. At the time, Musk was not yet an official part of the Republican presidential administration, but he had begun to make right-wing politics central to his personal mission, so in its infancy xAI was described as a creator of right-wing alternatives to AI products like ChatGPIT.
Last summer, users learned that Grok was willing to praise Hitler and advocate Nazi ideas, as my colleague Matt Novak cataloged in detail. xAI eventually issued a lengthy apology, saying it had falsified the internal workings of Grok.
Late last year, XAI’s signature chatbot, Grok, began to be used by X users to easily AI-edit images of other X users without requiring their knowledge or consent. The integration of Grok into X made it uniquely seamless, requiring only a single post directed at Grok. However, many of these images contained sexual material and near-nudity, and many of them depicted minors. The situation prompted regulators around the world to take various actions. XAI apologized for at least one image and said it “violated ethical standards and potentially US laws on CSAM.” It should be noted that the apology was prompted by an X user addressing Grok directly and apologizing.
Despite all the controversies, Musk seems largely satisfied with most of the company’s output in recent months. On February 17, he posted that the latest version of Grok, version 4.20, was “based” and included screenshots that showed that, compared to other popular chatbots, it refuses to even entertain the idea that America is “on stolen land.”
But if Musk has hinted at a change of direction for XAI’s actual products, it has nothing to do with the political leanings or troubling behavior of its core model. Instead, the Financial Times reported earlier this month that Musk was letting the co-founders go amid turmoil over xAI’s alleged failings around coding.
Appearing on a video call at an event called the Abundance Conference this month, Musk said he had just been in an all-hands meeting on coding, which was why he was late for that call. “He’s going through all the things we’re going through to try and outpace our competitors in coding, which I think we will. We should probably get there by the middle of this year,” he said.
Starting from 2025 to this year, XAI’s competitor Anthropic’s flagship large language model cloud is beginning to be considered an essential part of AI coding – especially coding and other coding-adjacent tasks performed through agentic AI platforms like OpenClause. Another of XAI’s competitors, OpenAI, has hired the creator of OpenClaw, and has turned to business and productivity applications for AI.
So xAI, in collaboration with Tesla’s computing division, is launching something, and is calling the venture “Macrohard.” Musk’s own descriptions of the software being produced by Macrohard are full of classic Musk bluster, and are unusually difficult to analyze for actual content. He says the product he has in mind will be “like a much more advanced and sophisticated version of turn-by-turn navigation software”, but that it will be “able to simulate the work of entire companies”. It appears that what he is referring to is a more capable version of OpenClaw.
As a major part of SpaceX during its IPO, Musk clearly wants investors to think big, which is why he’s hinting at plans to build Dyson spheres to absorb greater amounts of solar energy, build AI data centers in space, and soon surpass all combined human intelligence with AI. Most, if not all, AI company founders make big promises, but Musk’s promises might be the biggest of all.
What’s strange, however, is that the IPO plan would seem to contradict what Musk has said in the past about running public companies: He hates it. He has repeatedly said that he cannot afford the pressure put on him by shareholders expecting returns as CEO and the problems in the broader economy. The move would make X, formerly known as Twitter, once again part of a public company, as the entire purpose of buying Twitter was to take it private.
But to secure the top spot on the AI leaderboard, XAI will need to rapidly build energy infrastructure and data centers – and it is believed that this will take place literally in space. To that end, one possible purpose of the SpaceX IPO comes to mind: not only to be worth an estimated $1.75 trillion after the IPO and what that means for Musk’s net worth, but also for the company to net up to $80 billion from investors in the process. Much of the potentially record-breaking amount of investment will be a stockpile of spendable, liquid cash.
When you’ve made bad promises to the world, the only thing that can keep you from realizing it is a bad amount of money.
<a href