Young AI Founders Turn Down Musk, Unveil HRM Model

Young founders William Chen and Guan Wang, both in their early twenties and hailing from the Gen Z cohort, are making waves in the artificial-intelligence world. Their start-up, Sapient Intelligence, has developed a novel architecture dubbed the Hierarchical Reasoning Model (HRM), which they say marks a major departure from the typical large-language-model (LLM) paradigm.

Key Highlights

• Gen Z founders reject Elon Musk’s offer, developing HRM AI outperforming larger models in reasoning tasks.
• Sapient Intelligence’s HRM model shows advanced thinking, fewer hallucinations, and strong results in complex problems.

What is particularly striking is that they reportedly rejected an acquisition proposal from Elon Musk’s company xAI to instead continue building HRM independently. The decision demonstrates their ambition: they publicly state they aim to build an AI system that surpasses human-level reasoning rather than simply join a larger entity.

HRM itself leverages a two-module structure: a “high‐level” planner responsible for abstract, strategic reasoning, and a “low‐level” executor performing rapid detailed computations. Surprisingly, the prototype version of HRM uses only about 27 million parameters — a very small size compared to many mainstream LLMs with billions of parameters — yet, according to the founders and the published preprint, it outperforms much larger models on tasks requiring abstract reasoning, such as complex Sudoku puzzles and maze navigation.

They also claim that HRM suffers far fewer hallucinations (false or nonsensical outputs) than standard transformer-based models, and that early applications include domains like weather forecasting and quantitative trading. The founders note that if they do not venture into this territory, someone else will — hence their decision to press ahead. As Chen put it: “It’s not guessing… it’s thinking.”

Also Read: Nvidia CEO Slams Managers for Discouraging AI Use in Workplace

This story stands at several interesting intersections: young entrepreneurs in AI, architectural innovation beyond ever-bigger models, and a tactical decision to remain independent rather than being subsumed by a major tech-player. If successful, HRM could represent a shift in how we think about AI reasoning and efficiency. At the same time, its practical scalability and performance in diverse real-world settings remain to be established.