# DeepSeek Founder Liang Wenfeng Shares Key Insights from 4-Hour Investor Meeting > Source: > Published: 2026-07-23 02:46:15+00:00 Last month, «elsewhere» reported on [DeepSeek’s funding story](https://mp.weixin.qq.com/s?__biz=MzYzNTA3NTgwNQ==&mid=2247484219&idx=1&sn=35ba674230d6b7c355fb0e5acf940307&scene=21#wechat_redirect). The most widely discussed aspect was undoubtedly the legendary 4-hour investor meeting. Over the past month, various quotes from Liang Wenheng have circulated in the community, and we have gathered several of them. During this time, Liang Wenfeng said “no” many times: not a genius, not profiting from unreasonable gains, not pursuing user numbers, not going closed-source, not building 3D/video generation or world models, and not creating the next super app. He said that restraint is a strategy to increase the probability of achieving AGI. Among the limited materials we have reviewed, the following high-frequency terms appear: model, cost, AGI, time, open source, and others. Most of the time, Liang Wenheng speaks with restraint, using simple and unadorned language. Only when discussing a few matters he truly cares about does a hint of sharpness emerge: “As long as we maintain team stability, I will definitely achieve AGI. It’s that simple.” Here are the 52 statements we collected. Some phrases may differ slightly from the original while preserving the original meaning. ## DeepSeek has only one main line. ## 1. Now is not the time to maximize returns through products. On the path to AGI, products are a stepping stone, but we don’t need to invest excessive thought or effort into C-end or B-end products. When you operate from a high technological vantage point to develop relatively lower-level technologies, it’s a form of dimensional reduction. Products are byproducts along the journey to AGI. 2. Many things are not on our main path, such as 3D and video generation; likewise, world models are not closely related to the upper limits of intelligence. 3. Multimodality is important for the product and for end users, but it is only one component, not the core or the intelligence itself. 4. There are certainly ways to address hallucinations in large models, but this is a complex topic. Internally, we classify hallucinations as a product issue that we will resolve, but it is not our priority. 5. At this stage, the most important thing is the Coding Agent. Considering the situation in China, the most reasonable approach is to focus fully on developing a general-purpose Agent, with lower priority given to Agents in finance, healthcare, and other areas. 6. If the AI era gives rise to many trillion-dollar companies, it would be great if DeepSeek is one of them. ## Achieve continuous learning first, then AI self-iteration, with the end goal being embodied intelligence. 7. AI no longer lacks taste and intuition; it lacks the ability to learn continuously. 8. Humans can learn continuously, but for AI to understand the same thing, it needs all the context at once—something that is nearly impossible. Therefore, AI cannot replace employees. So, the next generation of models must have continuous learning capabilities to be called the next generation. 9. We hope the next-generation model will help our own development. Simply put, the primary goal of the models we build is not to be user-friendly for everyone, but to be highly effective for our own use. This is the fastest path to achieving AGI. 10. Currently, the world has not yet found a good method, as "learning" is composed of many elements. 11. DeepSeek’s long-term vision is AGI. If we compare the path to achievement as climbing stairs, last year’s step was CoT (Chain-of-Thought), and this year’s step is Agent. The challenge after Agent is continuous learning. 12. After implementing continuous learning, a gradual singularity may be reached: the model can perform all tasks that humans can do, including developing more advanced AI models on its own—that is, AI can accelerate AI research. Only after this step is completed will embodied intelligence be achieved. 13. The ultimate endpoint of intelligence may all be embodied. Because for an ordinary person, their need is not a computer, but human labor. ## Still far from a complete shift to commercialization 14. We charge a reasonable profit, not a profit-maximizing price. 15. One of our models initially had concerns about high demand and was priced accordingly, but we later reduced the price to a quarter of its original value, prompting many cheers in the company group. This was exactly our goal in putting so much effort into developing this model: to make it accessible to everyone. 16. Low cost is a result of our model continuously evolving toward a more cost-efficient architecture. We also aim to ensure that costs remain affordable, especially given the current scarcity of computing power. Another reason: the lower the cost, the larger the model we can afford. With limited computing power, higher computational efficiency enables training larger models. Large companies can solve this by adding resources, but we prioritize cost efficiency. 17. It might look from the outside like we chose a difficult model, but in reality, we made it very easy. Price cuts are certainly not good for our competitors—they’re not cheering. I don’t find selling APIs that attractive. I just need a few people to maintain the API; we don’t even need customer service or sales—users will come on their own. 18. We have been commercializing, but commercialization is not our goal. DeepSeek is still far from fully transitioning to a commercial model. 19. I don’t even need to worry about securing a spot inside; with such a massive business opportunity, there will surely be a way. DeepSeek is a product of its time, a reflection of current realities, not a copy. ## Open source is our company’s sweet spot at this scale. 20. Restraint is a strategy: giving up some to gain more of something else. Open source is about sharing value: internally, employees feel a sense of accomplishment and the company builds cohesion; externally, it benefits society, bringing joy to peers and the public alike. I have no doubt that AGI will have tremendous commercial value. On this basis, my priority is not to secure a larger share, but to increase the likelihood of success. 21. To make AI commercially successful, open source has advantages. This may sound counterintuitive. Historically, a software company might generate a few billion dollars in revenue in a year, and open-sourcing it would eliminate that. But AI is large enough that, in the end, it could account for 10% of human society’s GDP. If we try to monopolize this benefit, we will inevitably be abandoned by history. This is an objective law, a historical perspective. 22. The open-source model we provide is the same as the one we deploy ourselves. We do not open-source an inferior model while using a better one for our own deployment. 23. I’m not worried about others deploying our model to compete with us. Not every company has the willingness or capability to achieve this. Startups that are too small lack the power to do it, while large companies struggle to organize effectively. This is a sweet spot for a company of our size. 24. Open source has no impact on our business model as long as we aim to make only a certain profit. If you want to make a hundredfold profit, open source will indeed affect you. 25. We don’t want to become competitors with any large or small internet company. Under this premise, we are very willing to assist and support anyone, including Alibaba, Zhipu, and Moonshot AI, to do better. ## The gap between China and the U.S. is not in talent. 26. In the future, we will rewrite the narrative of AI in China and the U.S.: using a fraction of the computing power to narrow the gap to just six months, or even three months. 27. The gap between U.S. and Chinese AI primarily lies in resources. We believe in scaling—larger scale definitely leads to better results. We train models this large not because we think this size is sufficient, but because these are the resources we have. 28. There is almost no gap in talent—it’s the same group of people. There is no shortage of talent domestically. Talent shortages are temporary; historically, there has never been a long-term lack of any specific type of talent. ## Factory competition, cost comes first 29. Anthropic is currently surpassing OpenAI, not permanently, but temporarily. OpenAI and Google are likely to alternate in growth in the future. 30. There are too many companies in China building models, and each one is doing the same thing, leading to fragmented resources. Consolidation is inevitable, but it will take time. If each company takes only reasonable profits, far fewer people would be needed to develop large models: perhaps just two large companies and two small ones would suffice. I absolutely don't believe that large model companies can take the majority of profits (from the AI industry). 32. The ultimate differences in the large model competition will be reflected in three aspects: cost, time, and user experience. Cost comes first—what cost can you offer for the same quality of service? Second is time—being a few months earlier or later makes a difference. User experience may have some stickiness and barriers, but they are not essential. ## Not aiming to become the next super app 33. We don’t want to become the next super app. The idea of becoming the next ByteDance or the next Tencent? We have absolutely no such intention. 34. I’m not fighting over this because there’s a watermelon waiting behind it; what’s in front might just be sesame seeds. Sure, the sesame seeds might be relatively large, but I still don’t consider them big. 35. Last year, everyone was competing for chatbots and C-end traffic; this year, it’s about grabbing B-end revenue. But we don’t think it’s important—what truly matters inside the company is the AGI roadmap and how to achieve the next technological breakthrough. It’s strange: the things you want most are often the hardest to get, while the things you care less about come more easily. 36. Last year’s Spring Festival surge was not in our script. ## Maintaining team stability is essential. 37. There is only one thing that cannot be compromised: maintaining team stability. This is also a major risk we faced. Of course, this risk has been significantly mitigated with this round of funding. 38. We do many things to maintain team stability. We don’t want to become competitors with any large or small internet companies—we aim to empower them and help them. We don’t want to create enemies, because doing so will also create a better environment for ourselves. 39. Some people think our organization is top-down, while others think it’s bottom-up—I believe both are correct. The top-down approach is called “doing the right things,” and generally, “the right things” shouldn’t take up more than half of an employee’s time. The other half is bottom-up—unassigned, allowing individuals to explore whatever they find interesting, following their own sense of what matters, with no prior requirements. 40. We also don’t usually work overtime. The first reason is that research requires a relatively relaxed environment. The second reason is that we are highly focused. Many products are not fully developed, but we also choose not to fix them—that’s part of our culture of restraint. 41. Organizations are dynamic, not static. As the company grows in the future, there may be some adjustments. However, it will not fully become a traditional hierarchical structure, but some necessary structures may emerge. Yet, the vision-driven approach will remain unchanged. ## With goodwill toward the world 42. When we started this company, our original intention wasn’t to make a lot of money or go public on the capital market. The first dozen or so of us never even considered that. If anyone had thought that way, they wouldn’t have joined. We approached this with deep goodwill toward the world, believing it could be beneficial to humanity. 43. “Achieving XX KPI” is not our approach; we are a vision-driven organization. This has its advantages and disadvantages. We will find ways in the future to leverage the strengths and mitigate the weaknesses, but this is our defining characteristic. 44. The vision isn’t even formally written down—it’s reflected in how we do things and our attitude toward the world. Each person in our company may understand this vision differently, but we’re aligned on the big picture. About twenty years ago, my biggest management role model was Jack Welch, former CEO of GE. Today, much of what he said may no longer be relevant, but he was right about one thing: the most important thing for a company is its vision. Vision is not a slogan on the wall, not about what you say, but about what you do. ## Discipline helps us achieve AGI more effectively. 46. AGI offers the highest return. For other matters, if we have the energy, we’ll do them; if not, we won’t. Restraint is part of our vision. 47. AI is a huge matter with enormous potential profits. As long as you can pull it off, even a small share of the profits would be tremendous. The more restrained you are, the more likely you are to succeed. 48. I think this makes intuitive sense—at least to me. Beyond the vision, we don’t have many other advantages. 49. When we founded this company two years ago, we had little money, few cards, no recognition, and no influence—we were just a group of very ordinary people. I prefer the narrative of “a group of ordinary people achieving extraordinary things,” not “a group of geniuses achieving extraordinary things.” 50. Open source is also part of restraint. In terms of pricing, we certainly don’t start with the goal of maximizing company revenue or profit. In the short term, higher prices might mean more revenue; but in the long term, it’s truly uncertain. To me, restraint is a strategy. 51. Open source and low costs give employees a strong sense of achievement, foster organizational cohesion, benefit society, and bring joy to peers and the general public. Thus, this restraint increases, in the long term, our probability of achieving AGI. 52. If your vision is to take more, you’ve already lost—you may face even greater challenges. That’s just how the world works.