22-Year-Old Innovators Reject Musk's Millions to Build Revolutionary AI That Outperforms Industry Giants

Guan Wang and William Chen, co-founders of Sapient Intelligence, are revolutionizing the artificial intelligence landscape at just 22 years old.

These Michigan natives made headlines after declining a lucrative multimillion-dollar offer from tech magnate Elon Musk to instead pursue their groundbreaking AI venture.

The duo's friendship began in high school where they connected over ambitious "metagoals." Wang aspired to develop an algorithm capable of solving any problem, while Chen focused on optimizing systems across engineering and real-world applications.

"One day, we're going to have an AI that's smarter than humans," Chen explained to Fortune. "Guan and I always say it's like Pandora's box. If we're not going to make it, someone else will. So we hope that we're going to be the first one to make that happen."

After graduation, both attended Tsinghua University in Beijing. Despite initial academic challenges, they gained faculty support for their innovative AI research.

"We decided that large-language models have their limitations," Chen stated. "We want a new architecture that will overcome the structural limitation of [large-scale machine learning]."

Their initial breakthrough came with OpenChat, a compact large-language model trained on high-quality conversations designed for self-improvement through reinforcement learning. This project quickly gained academic recognition.

"It got very famous," Chen noted.

The model's success attracted Elon Musk's attention, leading to a substantial offer from his company xAI. Chen and Wang, however, declined the proposal.

This decision facilitated the establishment of Sapient Intelligence and the development of their Hierarchical Reasoning Model (HRM), which they claim outperforms major AI systems in abstract reasoning tasks.

Their breakthrough occurred in June when a prototype with merely 27 million parameters surpassed systems from industry leaders including OpenAI, Anthropic, and DeepSeek on complex challenges such as advanced Sudoku puzzles, maze-solving, and the ARC-AGI benchmark.

"It was crazy," Chen recalled. "Just with a change in the architecture, it gave the model a lot of what we call reasoning depth." Unlike conventional transformers that statistically predict the next word, HRM employs a dual recurrent structure mimicking human cognition, combining deliberate reasoning with quick reflexive responses. "It's not guessing," Chen emphasized. "It's thinking."

Chen further explained that their models produce significantly fewer hallucinations than traditional LLMs and already match industry-leading performance in weather prediction, quantitative trading, and medical monitoring applications.

Sapient Intelligence has announced plans to establish a US office in the near future.