When you start a company with 11 co-founders and end up with two, something has gone seriously wrong. That’s the reality at xAI right now, and Elon Musk isn’t even trying to hide it anymore. He’s basically admitted the whole thing was built wrong from the start and needs a complete reconstruction. Except the reconstruction is looking less like a careful rebuild and more like controlled chaos.
The latest casualties? Zihang Dai and Guodong Zhang, two of the original co-founders, just walked out this week after Musk complained that their coding tools were getting demolished by Claude Code and OpenAI’s Codex. An all-hands meeting happened Wednesday where Musk apparently spelled out exactly how badly they were losing, then promised they’d catch up by mid-year. Which, honestly, sounds like something you say when you’re panicking.
The Coding Tool Problem Is Actually a Money Problem
Here’s what most people miss about this mess: coding tools aren’t just a feature set. They’re where the actual revenue lives in AI labs. Sure, xAI got a bunch of early users excited about Grok because it would spit out sexual and abusive imagery with almost no guardrails. But that’s not a business model. Coding tools are. They’re what companies will actually pay for.
So when xAI’s coding tools lag behind the competition, it’s not some abstract benchmark thing. It’s the company failing at its primary cash-generation strategy. And Musk knows it, which is probably why he’s been aggressive enough to push out two founding team members in a single week.
The Exodus Won’t Stop
A month ago, 11 senior engineers left xAI following what Musk called a “reorganization.” That included two other co-founders. But apparently that purge wasn’t enough, because now SpaceX and Tesla executives have parachuted into the company to evaluate employees and identify who doesn’t meet the cut. That’s the kind of move you make when you’ve lost confidence in your current leadership structure.
The business problem is real: xAI has just over 5,000 employees compared to more than 7,500 at OpenAI and more than 4,700 at Anthropic. They’re smaller, they’re losing people, and they’re playing catch-up on the technology that actually matters.
Musk has started digging through rejected job applications with a colleague named Baris Akis, reaching out to candidates who didn’t make the cut before. He even apologized on X for ghosting people. That’s a pretty low moment for a company that’s supposed to be building the future.
Some Silver Linings, Sort Of
There’s one hiring win worth noting: Andrew Milich and Jason Ginsberg are joining from Cursor, an AI coding tool company. Their decision to move from a company that depends on frontier labs to access models over to xAI suggests something important. Direct access to your own frontier model and the compute to run it matters. A lot.
So maybe that’s xAI’s actual advantage here. They’ve got their own LLM. They’ve got their own compute infrastructure through SpaceX connections. That should theoretically be enough to compete.
Should be.
The Bigger Picture Nobody’s Talking About
The real tension isn’t just about coding tools. It’s about cash burn and investor confidence. xAI is now technically part of SpaceX, and there’s a SpaceX public offering anticipated. A stumbling AI division that’s bleeding cash and losing talent is not the story Musk needs circulating in investor meetings right now.
That’s probably why he’s been so public about the “rebuild” narrative. If the story is “we’re fixing this strategically,” it sounds better than “our AI lab is falling apart.” But the market’s watching employee exits, not press releases.
Macrohard: The Bigger Bet Nobody Asked For
Beyond coding tools, Musk is betting on something larger called Macrohard. Yes, that’s a intentionally funny reference to Microsoft, according to Musk. The goal is to build an AI agent that can do anything a white-collar worker can do on a computer.
Sounds great in theory. Execution has been rough. The project lead left within weeks. Business Insider reported it was on pause. Then this week, Musk revealed it’s actually a joint effort with Tesla, which is building a complementary agent called “Digital Optimus” (wink wink, humanoid robot reference).
The vision is ambitious but not original. Perplexity is already working on something similar. OpenAI has Peter Steinberger developing agents there. The technology space is crowded and moving fast, and xAI is trying to play catch-up while rebuilding its entire company culture at the same time.
The Real Question
Can a company rebuild itself while competing for market share against better-funded, more stable rivals? History suggests it’s possible, but it requires execution that Musk’s companies aren’t exactly known for right now. xAI has the assets. It has the talent pool available. It even has Musk’s attention and resources.
But attention from Musk might actually be part of the problem.
