Recently, I've been looking at some AI stuff, and the interview with Zhang Xiaojun and Zhu Xiaohu is quite interesting. Here are a few of his points summarized: 1. AGI has stagnated, and the scaling law based on the transformer architecture has failed. Both the US and China are starting to focus on applications. 2. What Sam says and what he does are different. He still talks about the AGI story, but in reality, OpenAI is working on applications, including the AI version of today's Toutiao Pulse and group chat features. Sam's past as an investor has made him more pragmatic. 3. Jinsha River Capital has always maintained a "pragmatic" investment style, disliking grand narratives. They prefer to deviate 15 degrees from consensus and invest in agents with vertical scenarios and commercialization capabilities. They haven't invested in large models or robots. 4. The food delivery war seems to be a lose-lose situation, which is meaningless. For Alibaba, the importance of Qwen and the new generation of AI super gateways is greater. Burning money on both ends is currently a significant pressure for Alibaba, and the coming period will test Alibaba's strategic choices. 5. Even if foundational models have started to stagnate, for many scenarios, fine-tuning the 3B Qwen is already sufficient. Token consumption is extremely fast, and consumers are not aware because many scenarios are B2B. He himself is most concerned with growth, retention, and commercialization. 6. The consensus among VCs is too strong now; it's all club deals, with good projects getting a small share from each firm. LPs are not very fond of this because it doesn't diversify risk, and GPs can't help LPs make big money since VCs rely on betting successfully on a big case and having a sufficiently large share to achieve excess returns. Zhu Xiaohu is still a VC investor with his own style. He focuses on scenarios and commercialization, preferring aggressive founders, which is why he invested in Ele.me and Didi. At the same time, he missed out on ByteDance and the "three electric vehicles" (Li Auto, NIO, Xpeng), with ByteDance alone representing half of the venture capital returns over the past decade. (You can find Zhang Xiaojun's account on major social media platforms; you can check out the full version yourself.)

What you think of the old Deng project team: made a lot of money, gathered a powerful team, unbeatable in operations, but in reality: they pulled a big one.

When you notice something is off, you need to cut your losses. I never expected the buyers/projects to have so much money. A few days ago, the 4%+ chips were all given to retail investors, and they all got caught... When logic conflicts with price, what you need to do in trading is to cut your losses and admit defeat first; never be stubborn. Holding onto a position is the dumbest thing you can do, so don't hold onto a position. Alright, I'm really not a forever-profitable influencer; this was a wake-up call. Fortunately, I cut my losses when I felt something was off yesterday. The possible explanations for the price being wrong are that the project team has money + the bears are too consensus-driven. My requirement for myself is to operate with 10 coins, and if I can win 7 of them, I'll be satisfied.