xAI lost two more co-founders this week—Guodong Zhang, head of the Imagine team, and Zihang Dai—over failures in the company’s AI coding product, leaving only 2 of the original 12 co-founders three years after Elon Musk founded the company in March 2023. Musk brought in “fixers” from SpaceX and Tesla to audit xAI, fired employees for inadequate work, and publicly admitted the company “was not built right first time around, so is being rebuilt from the foundations up.” The irony is brutal: an elite AI company that raised $45 billion and merged into SpaceX at a $1.25 trillion valuation ahead of a July 2026 IPO cannot build AI coding tools competitive with Claude Code, Cursor, or GitHub Copilot.
The Exodus Accelerates
Guodong Zhang told colleagues he was leaving after being blamed for issues with the coding product and relieved of his primary duties by Musk. Zihang Dai left earlier this week. These departures follow Tony Wu and Jimmy Ba (February 2026), Toby Pohlen (January 2026), and Greg Yang, bringing total co-founder attrition to 83% in three years. Financial Times sources said “the revolving door of talent was destroying morale.”
This isn’t normal startup churn. xAI’s founding team came from OpenAI, DeepMind, Google, and Microsoft—elite researchers who built foundational AI systems. When 10 of 12 co-founders leave within three years, the problem isn’t just product failures. It’s execution and culture breakdown.
Musk’s Admission: “Not Built Right”
Elon Musk publicly stated “xAI was not built right first time around, so is being rebuilt from the foundations up.” He brought in SpaceX and Tesla “fixers” to audit the company, fired employees for inadequate work, and complained that xAI’s coding tools were “not effectively competing” with Claude Code and OpenAI’s Codex. The response? Poach two senior Cursor engineers—Andrew Milich and Jason Ginsberg—to rebuild the coding division from scratch.
CEOs rarely admit fundamental failures this publicly, especially four months before a $1.5 trillion IPO. Musk’s admission reveals how dire the situation is. This isn’t product underperformance—it’s structural problems requiring complete rebuilds. Poaching Cursor engineers suggests xAI can’t solve this internally, even with $45 billion and access to every resource imaginable.
Related: AI Coding Tools Made Developers 19% Slower: METR Study
The Human Cost of “Extremely Hardcore”
xAI staff complained that management upheaval was “damaging morale and standing in the way of it reaching full potential.” Researchers left due to burnout from Musk’s “extremely hardcore” work demands or better offers from rivals. Former employees described a “culture of chronic burnout, rushed deployments, and sidelined safety practices.” An xAI engineer’s 19-hour workday went viral earlier this month, reigniting debate on hustle culture.
“Extremely hardcore” works at some companies. At xAI, it’s failing. When 10 of 12 co-founders leave and employees burn out publicly, the culture isn’t driving excellence—it’s driving talent away. This challenges Musk’s management philosophy: relentless intensity doesn’t guarantee results. Sometimes it backfires spectacularly.
Why Even Elite AI Companies Can’t Solve Autonomous Coding
Despite $45 billion in funding and an elite founding team from OpenAI, DeepMind, Google, and Microsoft, xAI’s coding products failed to compete with Claude Code (46% “most loved” rating among developers), Cursor (19%), or GitHub Copilot (9%). Grok Code Fast 1 scored 70.8% on the SWE-Bench Verified benchmark—technically competitive—but gained no market traction. The Financial Times reported Musk was “frustrated with coding product progress,” leading to co-founder firings and external audits.
This exposes the gap between AI coding hype and execution reality. If xAI—with unlimited resources and elite talent—can’t build competitive tools, maybe autonomous coding is fundamentally harder than the industry’s hype cycle suggests. Claude Code launched in May 2025 and became the number one tool within eight months. xAI released Grok Code Fast 1 earlier (August 2025) and lost the market. Money and talent aren’t enough when execution and culture fail.
Key Takeaways
- xAI lost 10 of 12 co-founders in three years (83% attrition) over AI coding product failures, morale collapse, and “extremely hardcore” culture—elite founding teams don’t leave unless there are serious execution and management problems
- Musk publicly admitted xAI “was not built right first time around” and is bringing in SpaceX/Tesla auditors to rebuild from scratch—a rare admission of fundamental failure four months before a $1.5 trillion IPO
- Despite $45 billion in funding and elite talent from OpenAI/DeepMind/Google/Microsoft, xAI’s coding tools failed to compete with Claude Code (46% “most loved”), Cursor (19%), or GitHub Copilot (9%)—revealing that autonomous coding is harder than hype suggests
- “Extremely hardcore” culture drove burnout instead of excellence—former employees reported “chronic burnout, rushed deployments, and reckless practices,” and the talent exodus proves relentless intensity can backfire
- xAI’s failure challenges the AI coding narrative: even with unlimited resources, elite companies can’t reliably build autonomous coding tools—money and talent aren’t enough when execution, culture, and product strategy fail
