This weekly roundup thread is intended for all culture war posts. 'Culture war' is vaguely defined, but it basically means controversial issues that fall along set tribal lines. Arguments over culture war issues generate a lot of heat and little light, and few deeply entrenched people ever change their minds. This thread is for voicing opinions and analyzing the state of the discussion while trying to optimize for light over heat.
Optimistically, we think that engaging with people you disagree with is worth your time, and so is being nice! Pessimistically, there are many dynamics that can lead discussions on Culture War topics to become unproductive. There's a human tendency to divide along tribal lines, praising your ingroup and vilifying your outgroup - and if you think you find it easy to criticize your ingroup, then it may be that your outgroup is not who you think it is. Extremists with opposing positions can feed off each other, highlighting each other's worst points to justify their own angry rhetoric, which becomes in turn a new example of bad behavior for the other side to highlight.
We would like to avoid these negative dynamics. Accordingly, we ask that you do not use this thread for waging the Culture War. Examples of waging the Culture War:
-
Shaming.
-
Attempting to 'build consensus' or enforce ideological conformity.
-
Making sweeping generalizations to vilify a group you dislike.
-
Recruiting for a cause.
-
Posting links that could be summarized as 'Boo outgroup!' Basically, if your content is 'Can you believe what Those People did this week?' then you should either refrain from posting, or do some very patient work to contextualize and/or steel-man the relevant viewpoint.
In general, you should argue to understand, not to win. This thread is not territory to be claimed by one group or another; indeed, the aim is to have many different viewpoints represented here. Thus, we also ask that you follow some guidelines:
-
Speak plainly. Avoid sarcasm and mockery. When disagreeing with someone, state your objections explicitly.
-
Be as precise and charitable as you can. Don't paraphrase unflatteringly.
-
Don't imply that someone said something they did not say, even if you think it follows from what they said.
-
Write like everyone is reading and you want them to be included in the discussion.
On an ad hoc basis, the mods will try to compile a list of the best posts/comments from the previous week, posted in Quality Contribution threads and archived at /r/TheThread. You may nominate a comment for this list by clicking on 'report' at the bottom of the post and typing 'Actually a quality contribution' as the report reason.
Jump in the discussion.
No email address required.
Notes -
I'll respond in more detail later, but for now I'll share a translation from a Zhihu post because it addresses several points here in an amusing way.
Who is Liang Wenfeng, the Founder of DeepSeek?
Qingfeng Xuezha (The Breeze Academic Underachiever) North American Computer Science Professor; Technological Innovation and Entrepreneurship; Providing Value, Emotional, and Knowledge-Based Services. Navigator Duan Xiaocao and 4173 others agree.
I've seen a lot of discussions about Liang Wenfeng online. Yesterday, I happened to have a phone call with a close friend from the same university year, and we also talked about Liang Wenfeng. So here I am, brazenly invoking my university classmate Liang Wenfeng. Some netizens want to know what Liang Wenfeng was like during his undergraduate days before he ventured into investment and the AI industry. This answer is meant to satisfy a bit of everyone's curiosity. I hope these "revelations" won't affect Liang's privacy. If they do, please remind me promptly, and I will modify or delete the answer.
The answerer and Liang Wenfeng were both in the 2002 cohort of Electronic Information Engineering at Zhejiang University (ZJU), not in the same class, but participated in the same Electronic Design Competition. Although we had some contact during the four years of university, because we weren't in the same dormitory or class, my impressions of Liang Wenfeng are limited and fragmented.
Impression 1: In our sophomore year, while we were obediently attending classes, doing homework, and preparing for exams, Liang Wenfeng was already self-studying digital circuits and analog circuits and had begun his own engineering practice. What left a deep impression was that he personally handled everything from circuit design, PCB layout, microcontroller programming, to UI design, creating something like a miniplayer software (doing Software UI in 2004 was a high-skill endeavor). He modified an ordinary guitar into an electric guitar, where the guitar's string sounds could be controlled via a UI on the computer. This project seemed incredibly impressive at the time; we all looked at it in awe. He humbly said the guitar's tuning wasn't great and it would be better if it could tune itself automatically. This can be considered a testament to the seed of his ideas about AI intelligence back then.
Impression 2: He rarely attended classes; most courses were self-taught. The answerer speculates the reason was he felt the teachers' pace was slow, a waste of time, and self-learning was faster. The downside was not following the teacher's emphasis on key points, which could hurt during exams. Liang Wenfeng's GPA in the major back then wasn't outstanding; it was upper-middle, not reaching the line for guaranteed postgraduate admission (保研线) (at ZJU back then, the proportion for guaranteed admission to the university's own postgraduate programs for ordinary majors was the top 5%). He later secured guaranteed postgraduate admission through winning the National First Prize in the Electronic Design Competition. This will be mentioned below.
Impression 3: During university, Liang Wenfeng traveled around several provinces in East China on his bicycle. Surprisingly, he often spent nights finding a spot in the wild to sleep on the ground, completing the trip without spending much money. This matter hasn't been verified; the answerer learned about it from the hot post "Liang Wenfeng, the Pride of 02 Telecommunications" on the 88 forum during graduation. The poster back then was also one of his teammates from the Electronic Design Competition, so the credibility should be quite high.
Impression 4: Liang Wenfeng and two other classmates from the same department signed up for the National Undergraduate Electronic Design Contest during the summer of their junior year. None of the three were top students in terms of academic grades, but their competition performance was outstanding. Naturally, Liang was the main force of the team. During ZJU's internal training camp, he single-handedly completed many design tasks. In the final competition, their team won first place in the province and the National First Prize. All three earned the qualification for guaranteed admission to ZJU's postgraduate programs without examination (免试推荐). However, because the national award announcement for the Electronic Design Contest that year was in October, they missed ZJU's guaranteed admission timeline for that year. Therefore, Liang could only start his postgraduate studies one year later. This explains the one-year gap between his undergraduate (2002-2006) and postgraduate (2007-2010) studies. It is said that during this gap year, he continued working on electronic sensing system design and products, something related to marine navigation, handling hardware, software, and algorithms all by himself. Every electronic system he built during his undergraduate years could easily suffice as a master's thesis for an electronics major.
Impression 5: Liang Wenfeng has always been low-key, just like he was during undergraduate days, so much so that many classmates in the same major weren't very familiar with him. Many heard of him through the National First Prize he won in his senior year. Therefore, it's not surprising to us, his university classmates, that he didn't come out to publish an article, say a word, or record a video amidst the overwhelming popularity of DeepSeek earlier. Ordinary people don't possess such composure and steadiness. (Addendum: Thinking back now, Liang Wenfeng isn't deliberately low-key; rather, his incredibly strong focus on his work makes him appear low-key - like Huang Yaoshi's final evaluation of Zhou Botong: "Old Urchin, Old Urchin, you are truly remarkable. I, Huang Laoxie, am indifferent to 'fame.' Master Yideng sees 'fame' as illusory. But you, with a mind empty and vacant, never had the notion of 'fame' in the first place, which puts you a step above us.")
Conclusion: Liang Wenfeng created his own success in his own way. He didn't live his university life according to the traditional standards of a "good student," nor did he study worldly social skills. He is a classic case of "Be Yourself" among Chinese university students and an example of contemporary intellectual youth entrepreneurship changing their own destiny (even the nation's destiny). Huanfang (幻方) was just the appetizer; DeepSeek is only the beginning. As an old classmate, I'm very happy to see him making outstanding contributions to the world's technological development and also honored to have seen the fledgling eagle before it soared across thousands of miles.
I hope the above sharing can provide some inspiration and motivation for China's tech-savvy youth. Chase your dream, and be yourself!
Yixiao Daxia (Smiling Hero) History and Current Affairs Enthusiast, Secretly Observing the World. 4358 people agree with this answer.
The answers are very fragmented. I carefully collected some information to try and organize it.
1. Birth Background and Early Experience
Liang Wenfeng was born in 1985 in Mili Ling Village, Qinba Town, Wuchuan City, Zhanjiang, Guangdong. His family circumstances were indeed ordinary; both parents were primary school Chinese language teachers, basically with no significant background. Liang Wenfeng made it mainly through studying.
Liang Wenfeng attended Meiling Primary School near his hometown in Wuchuan for elementary school. Both his junior and senior high school were at Wuchuan No.1 Middle School. He had some talent in mathematics; during junior high, he had already self-taught high school mathematics and started reading university-level math textbooks. In the 2002 college entrance exam (Gaokao), Liang Wenfeng scored 806 points, ranking first in Wuchuan No.1 Middle School, 14th in Zhanjiang City, and around 100th in Guangdong Province that year.
His first-choice application was for the Electronic Information Engineering major at Zhejiang University, graduating with a bachelor's degree in 2006. The year after graduation, 2007, he entered ZJU's Communication Engineering postgraduate program, graduating with a master's degree in 2010 (if it were a continuous bachelor's-master's program, graduation should have been 2009. Whether it was because he took the exam twice or something else is currently unknown).
2. Liang Wenfeng's Stock Market Life
As mentioned above, Liang Wenfeng had some talent in mathematics, and his undergraduate major was Electronic Engineering. Combining these two, the best application field he discovered was undoubtedly stock market trading. Therefore, during university, he developed a strong interest in financial trading. In 2008, Liang Wenfeng was 23, likely in his second year of master's studies, and began experimenting with automated trading in the A-share market with a principal of 80,000 RMB.
In 2010, the year he graduated, the stock market was in a downturn. However, it is said that Liang Wenfeng, through partly automated trading strategies, made 1 million RMB, gaining significant fame at the university and being called the "Campus Stock God."
After graduation, Liang Wenfeng did not seek employment nor start a business. He remained a retail investor, tinkering in the A-share market, continuously trying to write quantitative, automated strategies, testing them in the market to see if they could generate returns. It is said he once left Hangzhou and rented a place in Chengdu, closing himself off to trade stocks for three years.
If this stock trading venture hadn't succeeded, Liang Wenfeng would have been a typical negative example criticized by many. Imagine, a graduate from a prestigious university, not pursuing a proper career, and stock trading easily criticized as having a gambling addiction.
It wasn't until 2013, presumably after making considerable money from the stock market, that he began to end his status as an unemployed retail investor and started institutionalizing himself.
That year, he and his classmate Xu Jin established Hangzhou Yakebi (雅克比) Investment Management Co., Ltd. Generally, such asset management companies issue private fund products, get registered, and then raise money for investment. However, I guess it's likely that during the Yakebi phase, Liang Wenfeng and Xu Jin were similar to their previous retail investor status, probably lacking the qualifications and fundraising ability to issue products. The difference was having a company identity; their main work still focused on continuously researching, refining, validating, and improving their quantitative trading strategies.
After two years, the Hangzhou Yakebi company might have encountered issues, or perhaps they wanted to become a sunshine private fund (阳光私募), and the company didn't meet certain requirements, so it was abandoned. In 2015, he and Xu Jin together established a new company, Huanfang (幻方) Technology, and began the process of becoming a private fund manager (奔私).
In 2015, a recruitment post by Huanfang on Tsinghua University's Shuimu Community stated that Liang Wenfeng personally grew his 80,000 RMB principal from 2008 to 100 million RMB in profits over 7 years. It's unknown if this is true. If true, that's 1250 times in 7 years, basically tying with "Beijing Trader" as one of the fastest money-makers among retail investors in A-shares, and should be the domestic stock market's return champion. If the 100 million was accumulated through profit sharing during the Yakebi phase by raising significant external funds, then it involved substantial external leverage.
The period 2015-2017 was likely the most critical phase for Liang Wenfeng's stock trading. During this stage, all the quantitative trading explorations accumulated earlier finally bore fruit, and he successfully transitioned to a private fund institution, with asset management reaching a certain scale.
In 2016, Huanfang launched its first complete AI strategy. In 2017, they fully AI-ized their investment strategies. Presumably, their high-frequency trading AI strategy fit the characteristics of the A-share market very well, performing excellently. In 2017, Huanfang Quantitative's assets under management (AUM) broke through 30 billion RMB, and in 2018, they won the Private Fund Golden Bull Award (私募金牛奖).
Then things took off uncontrollably. In 2019, their managed funds exceeded 100 billion RMB. In 2021, they broke through 1 trillion RMB. However, by the end of 2021, perhaps due to the sheer size, over 100 products under Huanfang Quantitative saw declines exceeding 10%, causing investor losses. Subsequently, Huanfang Quantitative gradually reduced its funds under management.
By the end of 2024, Huanfang Quantitative's AUM was 45 billion RMB, with 63 fund products under its umbrella. However, performance differentiation is noticeable; 29 stock quantitative long-only products mostly maintained slight profits, while all 36 quantitative hedge-type products incurred losses. Of course, this is also related to the 2024 market conditions and policies. In 2024, the China Securities Regulatory Commission (CSRC) imposed significant restrictions on quantitative trading, likely preventing their high-frequency products from functioning normally.
So, here I must also advise my fellow A-share investors: you must work harder, diligently study the market every day, analyze companies, and focus on operations. Only then can you better compete on the same stage with Liang Wenfeng in the A-share market and defeat him.
3. Liang Wenfeng's AI Breakneck Advance
On October 21, 2016, Huanfang's first stock position generated by a deep learning algorithm model went live for real trading. They began using GPUs for computation. Before this, algorithms mainly relied on linear models and traditional machine learning algorithms, with model computation primarily depending on CPUs.
Since then, his breakneck advance in AI began. In 2019, Liang Wenfeng started large-scale procurement of GPUs, self-developing the Huanfang "Firefly One" (萤火一号) AI cluster, equipped with 500 graphics cards, interconnected with a 200Gbps high-speed network. In 2020, "Firefly One" had a total investment of nearly 200 million RMB, equipped with 1100 accelerator cards, and was officially put into use that year, providing computing power support for Huanfang's AI research. In 2021, presumably having really made money, Huanfang invested 1 billion RMB to build "Firefly Two" (萤火二号), equipped with about 10,000 NVIDIA A100 GPUs, breaking through the physical limits of the first phase and doubling computing capacity expansion.
After Huanfang's hardware and funding scale expanded, quantitative trading likely encountered some difficulties. Firstly, making money isn't as easy when the volume is too large. Secondly, the A-share market in 2023-2024 experienced a "Northern Myanmar"-like trend (a metaphor for a difficult/unpredictable market), with investors complaining bitterly, and regulators began supervising quantitative trading. Huanfang started reducing its funds under management from 2021, almost halving it. So, the hardware and computing power prepared for quantitative trading became idle and needed a new direction.
In 2023, Liang Wenfeng recognized the prospects in the field of Artificial General Intelligence (AGI). In July, he officially founded Hangzhou DeepSeek Artificial Intelligence Basic Technology Research Co., Ltd. (深度求索), focusing on the research and development of AI large models. In less than a year, in May 2024, DeepSeek released the mixture-of-experts language model DeepSeek-V2. On December 26, they launched and open-sourced the DeepSeek-V3 model, the version most of us used during the Spring Festival. The entire training process used less than 2.8 million GPU hours, costing about 40 million RMB.
On the evening of January 20, 2025, they released DeepSeek-R1. Its performance in mathematics, coding, and natural language reasoning tasks is comparable to OpenAI's o1 official version. They simultaneously open-sourced the model weights and training techniques, causing a huge stir worldwide.
DeepSeek directly shattered the American plan to monopolize cutting-edge AI technology and computing resources because it is both free and open-source. Anyway, I just made it casually; anyone who wants to use it can take it. I'm not making money from this, just for exploration and fun.
OpenAI: I'm getting a headache. I charge $150/month, $1800/year, and you're giving it away for free. What am I supposed to do? You're not charging either; what are you after?
Liang Wenfeng: It doesn't matter if I make money or not. What's important is that you can't make money!
I await your thoughts, not a copypasta of a chinese forum.
More options
Context Copy link
More options
Context Copy link