DeepSeek's boss Liang Wenfeng's internal speech to investors has been circulating online for a whole day,
and in the afternoon, it gradually got taken down from domestic public platforms.
I saw it was a document shared by group friends, but I found it quite intriguing!
The 42-page transcript discusses AGI, open source, commercialization, computing power, domestic chips, as well as organization management and talent.
There is a lot of information, but I think he repeatedly talks about only one thing: how to increase the probability of successfully achieving AGI.
For example, what impressed me the most is that DeepSeek chooses not to attack everywhere, but rather to continuously simplify.
Not in a hurry to grab users, not building the next super app;
Not for revenue, trying to capture everything in both the C-end and B-end;
Even if video generation, 3D, and world models are hot, it still doesn't follow;
Models can be closed-source to earn more but still choose to be open source;
Although prices could be set higher, it only earns what it considers reasonable profit.
This is not because he does not want to make money.
On the contrary, it is because he knows which money can be made, and which money earned now would slow him down.
1️⃣ Restraint is a strategy!
There is a saying that sticks with me: restraint is a strategy.
Many people's understanding of business competition is to grab users, seize entry points, and snatch revenue; whatever can be obtained is taken first.
But Liang Wenfeng’s understanding is completely opposite.
The field of AI is large enough that no single company can truly monopolize it. Since the subsequent opportunities are large enough, there's no need to stop for the immediate gains.
C-end users can be targeted, and B-end revenue can be earned, but these are merely by-products on the road to AGI and should not become the company's goals.
This is actually the most philosophical point throughout the entire discussion:
The more you want to gain, the harder it may be to achieve in the end; if you are willing to take a little less, you might find it easier to reach your goals.
Open source follows this logic as well.
Many believe open sourcing means giving up a competitive advantage, but Liang Wenfeng believes that as long as DeepSeek maintains advantages in cost and efficiency, open source will not only not harm business but can actually attract talent, build ecosystems, and reduce enemies.
He even suggests that if a company tries to take too much profit in the AI era, it will ultimately be defeated by another willing to take a little less.
Taking too little, the company cannot survive;
Taking too much will lead to challenges from new competitors.
In the end, the ones that can exist long-term are the ones with “reasonable profits.”
So his restraint is not a matter of being indifferent or having no desires.
Rather, it shows a strong sense of purpose:
Profits can be lesser, hot businesses may not be pursued, short-term users can be overlooked, but the two things that cannot be lost are: the main line of AGI and the stability of the core team.
This is where true restraint is sharp.
It’s not that he lacks desires, but that all his desires are focused on one thing.
2️⃣ Technical Roadmap
Technically, Liang Wenfeng has provided a clear roadmap:
After language models comes the thinking chain, after the thinking chain comes the Agent, and what really needs to be solved after the Agent is “continuous learning.”
Current AI capabilities are already strong, but it still requires humans to provide complete context. It cannot enter a company like a new employee, learn for two months, understand the people, relationships, rules, and work habits, and then continue to grow.
Once AI has the capability for continuous learning, it can start assisting humans in researching the next generation of AI, forming a cycle of “AI accelerating AI.” Beyond that, it will be self-iteration and embodied intelligence.
Of course, he also admits very honestly:
Continuous learning has not yet been truly solved, and the whole world is still exploring.
This kind of honesty is rare.
Believing that AGI will definitely happen while admitting that the specific path is still full of unknowns is not for financing but rather a firm stance.
3️⃣ The gap between China and the U.S.
Regarding the gap between China's and U.S.'s AI, his judgment is also straightforward:
The biggest gap is not in talent, but in computing power and resources.
In the context where a magnitude difference in computing power still exists, China cannot completely surpass others; it can only rely on higher efficiency, lower costs, and clearer trade-offs to catch up or even lead in certain directions.
He predicts that the real gap among future large models won't be some mysterious technology that can never be replicated, but three things:
Cost, time, and user experience.
Who can make it first, who can provide it at a lower cost, and whose product is more comfortable to use will be the one that stays.
4️⃣ Long-termism
The greatest inspiration I gained from this exchange is not which AI company to buy, but a re-understanding of the term “long-termism.”
True long-termism is not about claiming to have a far-sighted vision.
Rather, it is about whether you have the capacity to refuse when short-term interests are truly in front of you.
Investment is the same.
Trying to participate in every hot topic, seeking profit from every trend, fearing to miss out on every rise, ultimately causes funds, attention, and judgment to be scattered.
A person's true area of capability may not only involve knowing their strengths but also knowing:
Which profits simply do not belong to them.
However, Liang Wenfeng's model should not be oversimplified or romanticized.
DeepSeek's willingness to exercise restraint comes from its technological efficiency, its team, its funding, and the belief that the market ahead is large enough. Open sourcing, not setting KPIs, and letting employees explore freely cannot be replicated by just any company.
Liang Wenfeng himself acknowledges that as the company grows, some departments still require a clearer organizational structure; DeepSeek must also rely on APIs and commercial revenue to survive.
So what’s truly worth learning is not the superficial “not making money,” “not working overtime,” and “not setting KPIs.”
Rather, it's about first clarifying:
What is absolutely indispensable for oneself, and what can be voluntarily given up.
For DeepSeek, what cannot be lost is AGI and the core team.
For ordinary people, it may be health, family, cash flow, judgment, and the direction they truly wish to commit to long-term.
Today everyone is discussing how to have more, yet Liang Wenfeng spent nearly four hours explaining why he can afford to take less.
I think this might be the most powerful aspect of the entire exchange:
Being able to seize every opportunity is not what makes one impressive; instead, knowing where to truly go amidst countless opportunities is.
Many people believe the hardest part of long-termism is persistence.
In reality, the more difficult part may always have been letting go.
免责声明：本文章仅代表作者个人观点，不代表本平台的立场和观点。本文章仅供信息分享，不构成对任何人的任何投资建议。用户与作者之间的任何争议，与本平台无关。如网页中刊载的文章或图片涉及侵权，请提供相关的权利证明和身份证明发送邮件到support@aicoin.com，本平台相关工作人员将会进行核查。
