Insider Secrets of xAI | Generated by AI
Question: What did Sulaiman Ghori say in the podcast about xAI?
Answer:
Sulaiman Ghori, a former Member of Technical Staff at xAI, appeared on the “Relentless” podcast hosted by Ti Morse in an episode titled “WTF is happening at xAI,” published around January 15, 2026. In the roughly 1-hour-11-minute interview, he provided rare insider insights into xAI’s culture, operations, projects, and work with Elon Musk. Key points he discussed include:
-
Company Culture and Speed: xAI operates with extreme speed and high agency. Engineers have significant autonomy—no one tells them “no” if they have a good idea, which can be implemented and shown for feedback the same day. There is a flat structure with only three levels: individual contributors, cofounders/managers, and Elon Musk. Teams are small and elite (e.g., iOS team had 3 people, early Macrohard had 2). He emphasized radical autonomy, bottom-up innovation, rapid iteration (e.g., 24-hour cycles), and a “war room” surge culture with intense pushes, including weekends and long office hours.
-
Joining and Onboarding: Ghori joined xAI in March 2025 with minimal guidance—just a laptop and badge on day one. He worked on Ask Grok initially and jumped across projects.
- Major Projects:
- Macrohard: A key project to build “human emulators” or digital human simulators that replicate human digital behaviors (e.g., keyboard/mouse inputs, screen decisions) without needing specific software integrations. The goal is to automate any digital human task and deploy AI agents in real scenarios. They tested virtual AI “employees” internally, sometimes causing confusion (e.g., hallucinations leading to awkward interactions like asking coworkers to meet in person at an empty desk). Long-term aim: scale to millions of emulators, potentially challenging tools from companies like Microsoft.
- Grok and Related Features: Discussed Grokipedia (likely a truth-focused knowledge base for Grok), handling wrong Grok outputs seen by Elon on X (Elon intervenes directly), and figuring out “truth” in AI (describing it as a hard, chicken-and-egg problem tied to physics fundamentals).
-
Infrastructure and Hardware Leverage: xAI focuses on hardware/compute as the real bottleneck for AGI, not software. They built Colossus supercomputer quickly (122 days). Innovative approaches include using Tesla vehicle computers (especially HW4) as distributed compute (more capital-efficient than cloud VMs) for simulations and training—essentially “bootstrapping off the Tesla network.” Mentioned using mobile generators and battery packs for power balancing.
-
Elon Musk’s Style: Musk predicts bottlenecks far ahead, resolves issues rapidly (e.g., overnight patches), gives direct feedback, and is open to being proven wrong with data/experiments. Anecdote: Musk offered a free Cybertruck to an engineer (Tyler) if they could complete a GPU training run in 24 hours during setup—the engineer succeeded, and the Cybertruck is now parked outside the office.
- Other Insights: High per-commit value (~$2.5M), hiring via hackathons, challenging assumptions/requirements, experimentation over hierarchy, and using Tesla resources for macro hardware simulations to achieve scale (e.g., aiming for 1 million human emulators needing equivalent compute).
Ghori spoke positively about the high-trust, bureaucracy-free environment, though he left xAI shortly after (announced January 19, 2026, with “nothing but love” for the team), sparking speculation that the candid interview contributed to his departure (though unconfirmed).
References:
-
[YouTube: WTF is happening at xAI Sulaiman Ghori](https://www.youtube.com/watch?v=8jN60eJr4Ps) - Business Insider: 10 quotes about xAI and Elon Musk from the engineer who is out days after giving a sweeping podcast interview
- Times of India: Read the full interview that got engineer Sulaiman Khan Ghori ‘fired’ from Elon Musk’s xAI
- NextBigFuture: Inside xAI and Future of Millions of Emulated Humans