Source: AI Technology Review

This is the hottest technology star at the beginning of 2025. In just a few days, Liang Wenfeng All the past from childhood to adulthood have been shown to the world, including his new house that he had no time to decorate and the tent that he used to sleep in the house, which have become symbols of his unique personality.

Although unique personality is talked about, it is not the key to success. This unknown college student has relied on in the past decade only his thoughts and abilities.

Everyone is curious about this question, why did Liang Wenfeng make DeepSeek? There are certainly factors of the times and his personal experiences that are very different from other big model researchers. But AI Technology Review believes that understanding what kind of person Liang Wenfeng is is the key to understanding this issue.

Headhunter thinks that It is too difficult to find someone for Liang Wenfeng's company.

A headhunter who has worked in-depth with Fantasy since 2021 told us that recruiting people is so difficult that he can "want to cry" because it is too difficult.

"Tsinghua University's blog, six top-notch students, do you think there is no problem? Oh, why did you just die for your resume? A Tsinghua University undergraduate MIT PhD was eliminated in the second round of interviews."

If you want to find candidates within a large company, he believes that Huanfang and DeepSeek will basically not compete with companies, they will only compete with overseas companies such as Google and Meta.

Another headhunter couldn't help but feel overwhelmed when talking about DeepSeek, "I'm so picky. I pushed a young middle-level member with very good performance in Bytes, and I didn't know after talking. I was very surprised. , go and ask them, the answer I gave me is that this person has no passion for AI. People have done some AI Agent-related projects, and they usually don’t say such evaluations.”

Liang Wenfeng has no label for talents. Regardless of his academic background or past performance, he only looks at the person's personal ability and personal quality.

The extremely high talent threshold has created the DeepSeek today. Among the big model team, DeepSeek's talent thickness may not be comparable to that of top companies, but the talent density can definitely be said to be the first level.

In addition to DeepSeek's high salary, retaining these talents also fully respects creativity and idea management models. "No fixed team, no reporting relationship, no annual plan" is more trust than management. The book "Netflix Culture Manual" once said, "Excellent colleagues and difficult challenges are the biggest factors that attract people to work in the company." For AI practitioners, there is no more difficult than AGIA huge challenge.

To do the most difficult thing, you must find the best person and give sufficient resources and trust. Top talents who are trusted often bring great explosive power, and this theory can be confirmed in the rise of TikTok.

During the 2018 Spring Festival, Douyin added more than tens of millions of new users every day. A product manager in charge of growth once mentioned that this growth project has no performance pressure at all, and sent an email to the finance department, and his account had an additional budget of over 100 million yuan. He realized at the time, "What can you not win with such a team?"

DeepSeek is the same. The person whose resume is screened must be not a problem of education; if the interview is not passed, it must be a problem of ability; the needs of talents are concentrated in one sentence, is this person who can be trusted to work hard for AGI?

This is DeepSeek's view of talent. Understanding this view of talent is the first step to understanding Liang Wenfeng.

Although it has been done for many years Quantitative, but Liang Wenfeng does not think he is a finance person. His view of himself is, "I am an AI person, but I am a quantitative scenario."

Almost all those who have communicated with Liang Wenfeng said that he is a person who will not be disturbed by the outside world. "His way of thinking is extremely pure, and he pays special attention to the first principles", "and he speaks very slowly. ”, “As soon as you open your mouth, you hit the nail on the head.”

The characteristics of quantitative investment fit into his minimalist style—it does not need to deal with complex upstream and downstream industrial chains, but only needs to focus on pure market data.

To this day, Liang Wenfeng is still often immersed in his own technological world and focused on solving problems. For example, when doing a big model, he will tell others, "If you think it clearly, you can do it, as long as you have a card," and other difficulties are not within the scope of consideration.

The same is true for money. Money is used for investment or charity. As long as you can spend it in the right place, the losses are not worth mentioning.

At the end of 2023, there was a sign language model project aimed at helping deaf and mute people. Liang Wenfeng was found in order to attract investment. Liang Wenfeng proposed that the advantage of this project is its outstanding public welfare nature, and the disadvantage is that the market size is limited. The hidden danger is that this is a project of a top university college student team, and they may not persist for a long time.

Although it is very likely that he will not get any reward, he still proposes that as long as the team is willing to continue to advance the project, he will be willing to invest.

In the past, Liang Wenfeng would spend 500 million yuan every year on investment or charity, but now he spends this money on DeepSeek. Stock trading is for making money, and investing in big models is for AGI, that's all.

DeepSeek has nearly 20,000Ka, he is extremely generous to computing power. He promised to the above-mentioned sign language model team that the computing power cluster will be open to them at any time. But he is a little "stingy" and has high requirements for the utilization rate of these nearly 20,000 cards. He strives to be full and not idle.

These two behaviors seem contradictory, but if explained in a minimalist view, it will work: the card exists for use, and you can use it as much as you can, and don't waste it.

No one spent With the cost of spending money, DeepSeek's App has obtained 100 million users in just 7 days. What do Liang Wenfeng thinks about this miraculous growth?

A investor asked Liang Wenfeng this question specifically during the Spring Festival, but Liang Wenfeng seemed to have no concerns about such a large amount of traffic. The investors received the answer: "This is still a long way from AGI." ”.

This is not Liang Wenfeng pretending to behave. According to AI Technology Comments, DeepSeek only arranged two or three people to be responsible for App maintenance, dialogue web development, and recharge backend management. So it is normal that it is not easy to use.

DeepSeek's various deeds in the B-end market have been widely circulated. For example, their privatization deployment was priced at only 450,000 yuan, which not only included the right to use a H20 or 910b, but also included a large model service, with a service life of one year. At the same price, only 910b can be rented for one year on Huawei Cloud, which means that DeepSeek's big model is almost free to be provided.

Private deployment does not make money, and DeepSeek does not care whether it makes money by relying on APIs. An employee of a large company who is connected to DeepSeek complained that it has a "love to use but not to use" temperament, which is always difficult to use and always does not adjust.

No matter how large the customer or call volume is, it is not worth looking at. All major manufacturers have to queue up during peak hours, and the user experience is very poor. There are also a lot of feedback from major customers. DeepSeek is required to expand and then expand, and at least respond smoothly. Don’t fail once in two requests. This kind of failure is almost unbearable.

The outside world is noisy, but Liang Wenfeng doesn't seem to care much about this matter.

How should this situation be solved? Many companies are troubled by this. According to some internal rumor, Liang Wenfeng believes that major manufacturers are fully capable of finding ways to solve the problem of failure of requests, and they should provide themselves with the bottom line, rather than relying too much on DeepSeek to ensure services.

This answer makes people laugh.

It can be said that Liang Wenfeng now doesn’t care about all the possibilities of commercialization.

Today, when many teams are invested in the application, Liang Wenfeng once told a friend, "Don't keep watching."When you look at the application and industry, you can only confine yourself to the point of view, because when the time comes, everything you think is wrong now. And you put more time, energy and money into the wrong path. ”

This is a advice to friends and a practice of oneself. Putting energy into application and commercialization is a wrong path for Liang Wenfeng, no matter what he does.

There has never been one right path, and he is now on the right path.

p>