This duo turned down Elon Musk to build next-generation AI
30 Nov 2025

William Chen and Guan Wang, two 22-year-old friends from Michigan, have rejected a multi-million dollar offer from Elon Musk.

The duo are the co-founders of Sapient Intelligence and they met in high school in Michigan.

Their ambitious "metagoals" led them to create an AI that could solve any problem as per Wang, and optimize systems across engineering and real-world applications according to Chen.

Sapient Intelligence's journey and AI project
Academic pursuit

After high school, Chen followed Wang to Tsinghua University in Beijing.

Despite initial challenges with the rigorous curriculum, they got support from professors for their ambitious AI project.

They believed that existing large language models (LLMs) had limitations and sought a new architecture to overcome these structural constraints.

This led them to create OpenChat, a small LLM trained on high-quality conversations and designed to improve itself using reinforcement learning.

OpenChat's success and Musk's offer
Model recognition

OpenChat quickly gained fame in academic circles, catching the attention of Musk.

He approached the students through his company xAI with a multi-million dollar offer, which they turned down.

This decision led to the establishment of Sapient Intelligence and their new Hierarchical Reasoning Model (HRM), which outperforms major AI systems on abstract reasoning tasks.

HRM's breakthrough and unique architecture
AI innovation

In June, a prototype of HRM with just 27 million parameters outperformed models from OpenAI, Anthropic, and DeepSeek on complex tasks like advanced Sudoku puzzles and maze-solving.

Unlike traditional transformers that predict the next word statistically, HRM uses a two-part recurrent structure mimicking human thought.

This combination of deliberate reasoning with fast reflexive responses gives it what Chen calls "reasoning depth."

Sapient Intelligence's models and future plans
AI advancement

Sapient Intelligence's models hallucinate much less than conventional LLMs and already match state-of-the-art performance in weather prediction, quantitative trading, as well as medical monitoring.

The firm is now preparing to open a US office.

Chen and Wang see their work as a major step toward the next era of artificial general intelligence (AGI).