Two Michigan Grads Stun Tech World by Rejecting Elon Musk’s xAI Offer

Two 22-year-old entrepreneurs just turned down a life-changing offer from one of the world’s most famous billionaires. William Chen and Guan Wang rejected a multimillion-dollar recruitment package from Elon Musk’s artificial intelligence company, xAI. The pair made their shocking decision to focus solely on their own startup, Sapient Intelligence, and its revolutionary AI architecture.

The young founders believe current large language models are fundamentally limited. Instead, they are building a new type of brain-inspired system. Their tiny prototype recently outperformed giants like OpenAI and Anthropic on key reasoning tests, validating their bold gamble.

The Lucrative Offer That Couldn’t Compete With a Vision

The offer from xAI came after the duo’s first project, OpenChat, gained academic acclaim. Researchers from Berkeley and Stanford adopted their small, efficiently trained language model. This success put Chen and Wang on the radar of major players in the competitive AI field.

According to Fortune, who first reported the story, the financial terms were significant. For most young tech founders, joining a venture backed by Elon Musk would be a dream. Chen and Wang saw it as a potential distraction. They felt their path to artificial general intelligence (AGI) required full independence.

Their commitment was soon tested. In June 2025, their Hierarchical Reasoning Model (HRM) achieved a breakthrough. The model, with just 27 million parameters, solved complex puzzles that stumped much larger systems. This proved their architecture had unique potential.

How a Tiny Model Outsmarted the AI Giants

The benchmark results were undeniable. On the challenging ARC-AGI test, Sapient’s tiny HRM scored 40.3%. This beat OpenAI’s o3-mini-high (34.5%), Anthropic’s Claude 3.7 (21.2%), and DeepSeek’s R1 model (15.8%). The test measures abstract reasoning, a key hurdle for modern AI.

Their model also solved extreme Sudoku puzzles and found optimal paths through large mazes. These are tasks that require logical deduction, not just pattern recognition. Chen described the 3 a.m. moment they saw the results as “crazy.” The architecture itself provided what he calls “reasoning depth.”

The HRM design mimics aspects of human cognition. It uses a two-part recurrent structure for slow, deliberate thinking and fast reactions. This differs from transformers that predict text statistically. The founders argue this approach reduces hallucinations and enables genuine problem-solving.

Building the Future From Michigan to Beijing

Chen and Wang’s partnership began at Cranbrook Schools in Michigan. They bonded over a shared ambition to build an “algorithm that solves any problem.” After graduation, Chen made an unusual choice. He followed Wang to Tsinghua University in Beijing, often called China’s MIT.

It was in Tsinghua’s brain lab where their ideas crystallized. They immersed themselves in neuroscience and machine learning. Their thesis took shape: AGI requires efficient architecture, not just more data and computing power. This core belief led them to reject the conventional path represented by xAI’s offer.

Now, Sapient Intelligence is preparing to open a U.S. office. The founders are raising funds to scale their technology. They are exploring continuous learning, where AI systems absorb new information without full retraining. Their goal remains unchanged—creating a new foundation for general intelligence.

The story of these two young AI researchers rejecting Elon Musk is more than a career choice. It is a bet on a fundamentally different direction for the entire field. Their early success suggests that the future of artificial intelligence may be smaller, smarter, and built by those willing to say no to the biggest checks.

Thought you’d like to know-

Who are William Chen and Guan Wang?

They are 22-year-old co-founders of Sapient Intelligence. They met in high school in Michigan and later studied together at Tsinghua University in Beijing. They turned down a major offer from Elon Musk’s xAI to build their own AI company.

What is the Hierarchical Reasoning Model (HRM)?

HRM is Sapient’s brain-inspired AI architecture. It uses a recurrent structure for deliberate reasoning. The tiny 27-million-parameter prototype outperformed much larger models from OpenAI and Anthropic on abstract thinking benchmarks in June 2025.

Why did they reject Elon Musk’s xAI offer?

They believed joining xAI would limit their ability to pursue their unique architectural vision. According to Fortune, they wanted to overcome the structural limits of large language models, which required full focus on their own startup.

How did their small model beat OpenAI?

It scored higher on the ARC-AGI abstract reasoning benchmark. The HRM achieved 40.3%, compared to OpenAI’s 34.5%. The founders credit their novel architecture, which provides greater “reasoning depth” for solving logic puzzles and complex tasks.

What is Sapient Intelligence working on now?

The company is opening a U.S. office and seeking funding. They are developing continuous learning capabilities for their models. Their long-term goal is to create artificial general intelligence through efficient, brain-inspired design, not simply scaling up data.

When did this story become public?

Key details were reported by Fortune in late March 2025. The breakthrough benchmark results for their HRM model occurred in June 2025. The initial recruitment offer from xAI was made two years prior, while the founders were in Beijing.