Only 2 of 11 Co-Founders Left: xAI Hires Cursor’s Top Engineers to Rebuild Its Coding Tool from Scratch

Elon Musk publicly admitted this week that xAI “was not built right first time around” — and he’s not just talking about tweaks. The company is tearing down its AI coding tool and starting over, with two senior hires poached directly from Cursor, the startup that currently dominates the AI-assisted coding market with a $2 billion annualized revenue run rate.

The timing is brutal. Nine of xAI’s eleven original co-founders have now departed, including the people who were supposed to build the very coding tool that Musk now says failed. The remaining two — Manuel Kroiss and Ross Nordeen — are essentially the last ones standing from the team that launched xAI with Musk back in 2023.

The Cursor Connection: Who Are Andrew Milich and Jason Ginsberg?

On March 12, Andrew Milich and Jason Ginsberg announced they were leaving Cursor to join xAI and SpaceX, where both will report directly to Musk. Their mandate: rebuild Grok’s coding capabilities from the ground up.

These aren’t random hires. At Cursor, Milich and Ginsberg jointly held the title of Head of Engineering, Product, overseeing the entire product engineering operation. They joined Cursor together in June 2025, and during their tenure, the company hit that eye-popping $2 billion annualized revenue figure — achieved with a lean team of roughly 20 engineers shipping updates every two weeks.

Before Cursor, the two have a shared history that goes back years. They co-founded Skiff, a privacy-focused collaboration tool that grew to 2.5 million users before Notion acquired it. Milich went on to lead Notion Mail, while Ginsberg served as Skiff’s CTO. Both hold Stanford computer science degrees — Milich with an AI concentration, Ginsberg with a master’s. Ginsberg also has stints at Apple and Sequoia Capital on his resume.

In short, Musk didn’t just hire two engineers. He hired the duo that helped scale the most successful AI coding tool on the market.

Why xAI’s Coding Tool Failed

The backstory here matters. Grok, xAI’s chatbot, has consistently lagged behind competitors when it comes to coding tasks. Internal staff reportedly pointed to insufficient training data as a core issue — the model simply wasn’t competitive with Anthropic’s Claude Code or OpenAI’s Codex for real-world programming work.

Musk himself acknowledged the gap bluntly: “Grok is currently behind in coding.” But behind the public statements, the situation was more chaotic. Co-founders Zihang Dai and Guodong Zhang — both deeply involved in xAI’s technical direction — departed after Musk expressed frustration that the coding tools couldn’t keep pace with the competition.

There was also the “Macrohard” project, an ambitious initiative led by co-founder Toby Pohlen, a former DeepMind researcher. Pohlen’s departure came just 16 days after his appointment to lead the project. In a post afterward, he said his priorities were to “sleep for more than 8 hours, write down all the things I’ve learnt… and then think about what I want to do next.” That kind of statement from a departing co-founder speaks volumes about the internal culture.

The broader pattern is hard to ignore. Since January 2026, six co-founders have exited in rapid succession — Toby Pohlen, Jimmy Ba, Tony Wu, Greg Yang, Zihang Dai, and Guodong Zhang. Combined with earlier departures, only Kroiss and Nordeen remain from the original founding team.

The AI Coding Tool War: Where xAI Stands Against Cursor, Claude Code, and Codex

The AI coding assistant space has become one of the most fiercely contested markets in tech, and xAI is entering a rebuild at the worst possible time.

Cursor is the current market leader among professional developers. Its deep codebase understanding, fast iteration speed, and tight editor integration have made it the default choice for many engineering teams. With $2 billion in annualized revenue from a 20-person engineering team, Cursor’s efficiency is almost absurd — and losing two of its product engineering leaders to xAI is a real blow, though the company still has a massive lead.

Claude Code from Anthropic has emerged as a serious force in coding. Developers have praised its ability to handle complex, multi-file coding tasks with strong reasoning capabilities. It’s become particularly popular for its terminal-based workflow and its ability to understand large codebases in context.

OpenAI’s Codex remains a heavyweight, backed by the company’s massive resources and integration into the broader ChatGPT ecosystem. Codex benefits from OpenAI’s scale and brand recognition, though some developers have noted that Anthropic’s offerings have closed the gap or pulled ahead in certain coding benchmarks.

GitHub Copilot, powered by various models including OpenAI’s, maintains a large installed base through its VS Code integration and GitHub ecosystem. It was the first mover in AI-assisted coding and still holds significant market share through enterprise contracts.

Against this field, Grok’s coding capabilities have been described internally as a generation behind. The question isn’t whether xAI can catch up — it’s whether the Milich-Ginsberg rebuild can close a gap that the original founding team couldn’t.

The Bigger Picture: SpaceX IPO Pressure and Internal Turmoil

This isn’t happening in a vacuum. xAI merged with SpaceX, and the combined entity is targeting an IPO that could value the company between $1.25 trillion and $1.75 trillion. That creates enormous pressure to show that every division — including xAI — is performing.

Musk has responded to the crisis by bringing in “problem solvers” from SpaceX and Tesla to conduct audits and identify underperforming staff, leading to another round of layoffs beyond the reorganization announced a month prior. The management style has drawn internal criticism, with employees reportedly complaining about micromanagement and burnout.

The revolving door of leadership creates its own problems. When nine of eleven co-founders leave a three-year-old company, institutional knowledge walks out with them. No amount of talent acquisition can fully replace the context and vision that founding team members carry.

That said, Musk has pulled off improbable turnarounds before. SpaceX itself was on the brink of failure before its fourth Falcon 1 launch succeeded. Tesla nearly went bankrupt multiple times. The pattern of crisis-driven restructuring is familiar territory for Musk — though whether it works in the fast-moving AI space, where competitors are shipping meaningful improvements on a weekly basis, remains an open question.

What This Means for the AI Coding Market

The xAI rebuild signals something important about the state of AI coding tools: the market is maturing faster than most companies can keep up. Building a competitive coding assistant isn’t just about having a capable base model — it requires deep product thinking, tight developer workflows, and a rapid iteration cycle.

Cursor figured this out early. Their small team, fast shipping cadence, and obsessive focus on developer experience created a product that developers actually want to use. By hiring Milich and Ginsberg, Musk is essentially trying to transplant that DNA into xAI.

The risk is that what made Cursor work — a small, focused team with autonomy — is fundamentally incompatible with xAI’s current structure of top-down mandates, executive audits, and constant reorganization. Great developer tools tend to come from teams that deeply understand and empathize with developers. Whether that culture can thrive under Musk’s management style is the real question.

Musk has set a target of closing the coding gap by mid-2026. Given that Milich and Ginsberg just started, that’s an aggressive timeline — essentially three months to go from “not built right” to competitive with tools that have years of head start.

FAQ

How does xAI’s coding tool compare to Cursor?
As of March 2026, they’re not in the same league. Cursor has $2 billion in annualized revenue and is the most widely used AI coding tool among professional engineering teams. xAI’s Grok coding capabilities have been described by Musk himself as “behind.” The hire of two senior Cursor leaders is an attempt to close that gap, but the rebuild is starting from scratch.

Why did xAI’s co-founders leave?
Nine of eleven original co-founders have departed since xAI’s founding in 2023, with six leaving since January 2026 alone. Reports point to a combination of factors: frustration with the coding tool’s inability to compete, internal culture issues, management pressure, and burnout. Co-founder Toby Pohlen’s post-departure comment about needing to “sleep for more than 8 hours” suggests an unsustainable work environment.

What is the “Macrohard” project at xAI?
Macrohard was an ambitious coding-related initiative at xAI, led by co-founder Toby Pohlen. The project appears to have stalled following Pohlen’s departure just 16 days after being appointed to lead it. Layoffs have also affected parts of the Macrohard team.

Can xAI catch up to Claude Code and Codex in coding?
Musk has set mid-2026 as the target for closing the coding gap. With experienced hires from Cursor and resources from the SpaceX-xAI merger, the company has the talent and funding. However, competitors like Anthropic and OpenAI are not standing still — they’re shipping improvements weekly. The challenge isn’t just building a good tool; it’s building one fast enough to match a moving target.

How does this affect Cursor?
Losing two heads of product engineering is significant, but Cursor’s position remains strong. The company’s $2 billion revenue run rate, established developer base, and product momentum provide a substantial buffer. The bigger risk for Cursor is whether xAI’s deep pockets and Musk’s visibility could attract more of their talent over time.

Top AI Product