Former Engineer’s Interview Sparks Departure Controversy
In a podcast interview in January 2026, former xAI engineer Sulaiman Ghori disclosed a significant amount of internal information about the company’s core projects and operations. Shortly after, he announced his departure on social media, sparking widespread industry discussion about being “fired for leaking.”
“MacroHard”: An AGI in Pursuit of Speed
Ghori revealed a general intelligent agent project named “MacroHard” in the interview. This project is positioned as “the Optimus Prime of the digital world,” aiming to automate all human decision-making tasks that require a keyboard, mouse, and screen. Unlike the industry’s mainstream “Scaling Law” approach, which relies on large-scale models to enhance reasoning abilities, xAI has chosen a small model path where the core metric is speed. The requirement is that the model must execute tasks much faster than a human. It was revealed that the model has already reached 8 times human efficiency.
Compute Innovation: Envisioning a Tesla Distributed Network
The most disruptive information from the interview was xAI’s vision for computing resources. To solve the compute bottleneck caused by large-scale agent deployment, xAI is considering leasing the idle in-car computer resources of approximately 4 million Tesla vehicles in North America. This plan treats Tesla vehicles as natural distributed computing nodes, utilizing their built-in FSD chips, batteries, and network connectivity to build a massive “compute gold mine.”
“Lightning-Fast Culture” and Extreme Execution
Ghori described xAI’s internal culture as “lightning-fast.” The company has an extremely flat organizational structure where engineers enjoy a high degree of autonomy, encouraging rapid bottom-up innovation and iteration. This culture has led to astonishing execution efficiency, such as building the Colossus 1 data center in just 122 days. However, this high efficiency is accompanied by immense pressure, with intense overtime becoming the norm.
