xAI looked like it had everything going for it three years ago. Elon Musk, a star-studded founding team of 11, and the audacity to take on OpenAI and Anthropic in one of the most competitive spaces in technology. Now? Only two of those original co-founders are left standing.
This week, two more co-founders walked. Zihang Dai and Guodong Zhang departed after Musk complained that xAI’s coding tools weren’t cutting it against Claude Code and Codex. That’s not just a personnel shuffle. That’s a death knell for morale.
The thing about co-founders leaving is that it sends a signal. If the people who believed in this thing from day one are now heading for the exits, what does that tell the engineers still in the building? What does it tell potential hires?
The Coding Tool Problem
Musk’s public explanation was blunt: xAI was “not built right first time around” and is being “rebuilt from the foundations up.” Translation: we screwed up. At least he’s honest about it.
But honesty doesn’t fix the immediate problem. Coding tools are where the real money lives in AI right now. This isn’t some vanity metric. Companies will pay for AI that makes their developers faster. xAI’s lag in this category is a business problem, not a perception problem.
Musk promised they’d catch up by mid-year. That’s a bold claim when you just lost two co-founders who presumably knew how to build this stuff.
The Bleeding Doesn’t Stop There
A month before this week’s departures, 11 senior engineers walked, including two other co-founders. The stated reason was a reorganization for a “larger business.” But here’s where it gets interesting: SpaceX and Tesla executives have reportedly parachuted into xAI to evaluate and fire underperforming employees.
When Elon starts sending in the goon squad from his other companies, you know things aren’t going well internally.
xAI has just over 5,000 employees. OpenAI has 7,500. Anthropic has 4,700. Numbers matter less than momentum, sure, but you can’t afford to lose senior talent when you’re playing catch-up in a talent war.
A Small Win Among the Wreckage
There’s one encouraging sign: Andrew Milich and Jason Ginsberg just joined xAI from Cursor, an AI coding tool company. They were responsible for product engineering over there, which means Musk is actually hiring people who know how to build what xAI desperately needs.
Their decision to leave Cursor for xAI is interesting. Cursor depends on frontier labs for access to models. xAI owns its own frontier model. Maybe that direct access to LLMs and compute is enough of a draw to pull serious talent into a sinking ship.
Or maybe it just means they saw the opportunity before things get even messier.
The Macrohard Distraction
Then there’s the Macrohard project. You know, the one with the “funny reference to Microsoft” that Musk is convinced is hilarious. It’s supposed to be an AI agent that does white-collar work. The project lead lasted weeks before leaving. Then Business Insider reported it was on pause.
This week, Musk revealed Macrohard is actually a joint effort with Tesla. Digital Optimus, Tesla’s complementary agent, would work alongside xAI’s language model. It’s ambitious. It’s also not particularly novel. Perplexity is doing similar things with “Everything is Computer.” OpenAI has people working on this too.
When your moonshot project can barely keep leadership and your “pivot” is just matching what everyone else is already doing, you’re not disrupting anymore. You’re following.
The Clock is Ticking
xAI is now part of SpaceX. A SpaceX IPO is on the horizon. A cash-burning AI division that can’t compete in coding tools and can’t keep its leadership intact is not the story Musk wants investors reading in their morning briefings.
The pressure isn’t just internal anymore. It’s the kind of existential pressure that forces hard decisions.
What happens when a company built by leaving people admits it was built wrong? When it rebuilds, does it actually get better, or does it just become a different version of broken?
