I probably figured it out tonight Liang Wenfeng should have figured it out long ago Tencent may not have figured it out yet I remember that in the mobile Internet era, all companies were afraid of Tencent and Ali. Because traffic was king, they could succeed if they asked for a little traffic. They harvested all the companies in that era. But recently, Deepseek has been very strong in its financing and has stopped playing with them. Alibaba has been eliminated and only given Tencent a small share to pass the time. Why so dominant? Because things have changed, the era of traffic being king is coming to an end. It has become an era where AI big models dominate the killing line. The AI industry currently has four directions 1. Big model, this is the core. At present, the intelligence level of large models cannot be monopolized, and they will be replaced almost every two to three months. But cost-effectiveness can be monopolized through various optimization methods. For example, Deepseek Kill Line can achieve a maximum intelligence level of 80%, but its price is only 1/10, so API calls can become monopolized. Can lead in a cliff like manner. That's why Liang Wenfeng doesn't care about Tencent's traffic or C-end users. Because he has only one goal in mind, which is the universal artificial intelligence agi, which takes everything and is like a black hole. Make a cost-effective agi that can handle everything. Your massive traffic is useless. You also have to use me. I remember eight years ago in the digital currency industry, there was a popular phrase 'fat protocol, thin application', but now the industry is almost disappearing, leaving only a stablecoin super application. However, I think the term 'fat protocol, thin application' is quite suitable for the AI industry. The agi kill line is a fat protocol that takes everything and is extremely valuable. And app terminals have become worthless at all. What Liang Wenfeng wants is not C-end monthly life, he wants all applications to grow on the "foundation" of DeepSeek. This is the dimensionality reduction blow of the "fat protocol" on "thin applications". 2. Hardware. Needless to say, it's very profitable. From a pitiful electronics factory to a provider of human AI infrastructure. For a long time, there will be a shortage, at least until agi is implemented, especially in terms of storage. 3. Cloud vendors. It's pretty good now, very profitable. However, the risks are also significant. Because the hardware they bought was extremely expensive, their cash flow was negative. I have earned money from selling computing power, but in the future, my free cash flow may not be optimistic and I may need to raise funds, which poses a huge depreciation risk. It's a forced business. 4. Terminal applications. So Tencent hasn't figured it out yet. It's because we're still investing a lot of resources in creating a lobster like workplace, and these applications are transitional products. On the day agi is implemented, this thing will no longer be needed. Tencent is still using traffic thinking to make products, but in fact, things have changed. Liang Wenfeng is tirelessly working on the killing line, which is the most profitable. Don't worry about any traffic either. As long as the model can keep up with the top tier and the price is only one tenth, the market share will become astonishingly leading. Use the computing power of cloud vendors to process and sell tokens using their own large models. This is the best business in the era of AI. Next is hardware. Cloud vendors and terminal applications have become bad business. Of course, there is also an old fox, Apple has not yet invested in AI and is still watching. I will only invest in hardware and large models in the future.