What drives DeepSeek?
A dive into its low-profile founder, Liang Wenfeng.
What drives Liang Wenfeng, the low-profile founder behind DeepSeek?
An outsider to high tech circles. Eccentric. That no one took seriously at the start.
Born in “fifth tier city” Guangdong, China where fortunes were made in factories and street markets, few had time for books. His father, an elementary school teacher, believed education was the only wealth that lasted.
In 2008, while still a student at Zhejiang University, Liang started using AI for quant trading. By 2021, his fund, Huanfang Quant, hit $13B AUM—earning a spot among China’s “Four Kings” of quant trading. Yet, on the global stage, it was still dwarfed by giants like Bridgewater ($126B AUM).
Then, in 2024, amid market turbulence, he did the unthinkable—he closed shop and walked away from one of China’s largest quant funds.
To build something even bigger.
The pursuit of AGI.
His business partners thought he was crazy. This was a game for trillion-dollar tech giants—not a scrappy team of finance nerds.
But DeepSeek’s latest release—a reasoning model called R1—would shake the AI world, proving that efficiency could outmaneuver brute force.
It created a market shock that wiped nearly $600B from tech stocks in a single day. (More on this in my next post.)
So what drives Deep Seek? These are the five insights into how Liang built one of the most competitive AI research companies in the world.
Learn From the Interns
DeepSeek is built differently. It’s not filled with big tech veterans or Silicon Valley elites—it’s mostly young people.
“There are no wizards. We are mostly fresh graduates from top universities, PhD candidates in their fourth or fifth year, and some young people who graduated just a few years ago.”
Their hiring standard? Talent and curiosity. “Their desire to do research often comes before making money.”
📖 Source: ChinaTalk
Open Source Is the Real Competitive Edge
“Moats created by closed-source models are temporary. Even OpenAI’s closed-source approach can’t prevent others from catching up.
So we anchor our value in our team — our colleagues grow through this process, accumulate know-how, and form an organization and culture capable of innovation. That’s our moat.”
DeepSeek isn’t just betting on AI—it’s betting on people. In the pursuit of AGI, Liang believes it’s a matter of who has the best team.
📖 Source: ChinaTalk
The business models we know today are irrelevant
“All business models we know today are products of the past generation. Using internet-era logic to predict AI’s future is like comparing Tencent’s early days to General Electric or Coca-Cola—it’s pointless.”
Most AI startups scramble to commercialize. DeepSeek refuses to chase short-term profits. Instead, it’s focused on solving the hardest problems in AI—because true breakthroughs come from exploration, not monetization.
📖 Source: Stratechery
China Needs to Be an Innovator, Not Just a Follower
“Why is Silicon Valley so innovative? Because they dare to try. When ChatGPT debuted, China lacked confidence in frontier research. But innovation requires confidence, and young people tend to have more of it.”
For years, China’s tech industry has been seen as a follower, not a leader. DeepSeek wants to change that.
“What we see is that Chinese AI can’t be in the position of following forever. We often say that there is a gap of one or two years between Chinese AI and the United States, but the real gap is the difference between originality and imitation. If this doesn’t change, China will always be only a follower.”
📖 Source: Sina Finance
The Three Approaches to AGI
When will AGI happen? “It could be two, five, or ten years–in any case, it will happen in our lifetimes.”
DeepSeek is betting on three directions towards AGI:
1️⃣ Mathematics & Code – Verifiable systems for self-taught intelligence
2️⃣ Multimodality – AI that understands the world like humans do
3️⃣ Natural Language – The core of human-AI interaction
“We remain open to different possibilities.”
📖 Source: ChinaTalk
DeepSeek isn’t playing by the old rules. It’s proving that in AI, efficiency beats brute force, and open beats closed.
The “dark horse” in AI will be one to watch.