Podcasts

This episode features seasoned investor Zhuang Minghao, who reviews his career path since joining Shanda Investment in 2009, witnessing the rise of China's mobile internet and the development of the primary market. The program deeply analyzes information asymmetry in the early investment environment and how investors make decisions amidst this uncertainty, such as Shanda's 'uninformed investment' in companies on the list back then. The podcast highlights why giants like ByteDance and Pinduoduo were 'missed' by most mainstream investment institutions early on. Zhuang Minghao explains how investors' biases and path dependencies, shaped by past experiences, impact their judgment. He also introduces an innovative perspective, suggesting that investors are essentially large language models. He likens early-stage investment to a 'ritual for rain,' emphasizing the psychological need to find certainty in high-uncertainty environments. Additionally, the episode explores the role of luck in investment success, the formation of investment taste, the VC industry's attractiveness to young professionals, and the new challenges presented by the spillover of capital liquidity and the influx of state-owned capital under the current AI wave. Finally, Zhuang Minghao shares the driving force behind his long-term writing and his profound understanding of content value, providing listeners with in-depth insights into investment, business, and personal growth from multiple dimensions.

This episode of 'Unconventional Wisdom' invites Silicon Valley 101 researcher Yiming Liu to discuss two core topics in the AI industry: the release of GPT-5 and the logic behind it, and the impact of the explosive growth of AI computing power on the US power system. The podcast first reviews the significant role of Reinforcement Learning in AI model training, especially in the pre-training phase, believing that it has alleviated the 'bottleneck' problem of previous data training. Subsequently, the guests analyzed the reasons why the release of GPT-5 was not as expected but emphasized enterprise-level applications, pointing out that OpenAI is actively expanding the To B market to support its huge capital expenditure. Finally, the podcast deeply analyzes the surge in electricity demand caused by the construction of Data Centers, which not only promoted the strong growth of power stocks but also foreshadowed that the fragile US power infrastructure will face a comprehensive reconstruction and huge investment, from power generation, power transmission to power distribution will usher in profound changes, and may even become a US version of infrastructure project.

This podcast focuses on the intensely competitive AI Companion field. Guests Momo and Chen Tuimao share the concept and practice of their startup project, Starfy. The team aims to solve the problem of the lack of high-quality friendship during the confusing period of life for modern people (especially urban women aged 20-35) through AI technology, creating immersive and personalized AI friends, rather than traditional AI-powered social platforms or Virtual Idols. The podcast delves into the essential differences between AI Social, AI Companion, and AI communities. It also analyzes the inspirations and limitations of existing products such as C.ai. At the technical implementation level, the guests emphasized that obtaining high-quality corpora (created by people with high emotional intelligence and user contributions) and memory functions (building consensus similar to old friends) are core challenges. In terms of business models, the team prefers to achieve long-term value through IP Derivatives rather than subscriptions, and emphasizes the importance of choosing new IPs that are relatively untapped. Facing market competition, the team believes that they should focus on building unique user relationships and internal production systems, rather than blindly pursuing speed or imitating major players in the industry. The podcast also discussed the depth of understanding of user needs and cultural particularities in the domestic market, as well as the innovation challenges and internal games faced by large corporations in the AI Social field. Overall, the podcast provides a comprehensive perspective on AI Companion products from concept to implementation, with in-depth thinking on product design, technical challenges, and business strategies.

This podcast delves into the development and challenges of OpenAI's "Project Stargate" since the beginning of the year. The guests first reviewed the significant changes in market sentiment in the AI industry, noting that despite early concerns about high investment in data centers, the surge in NVIDIA's stock price indicates the continued growth in AI computational power demand. They also warned that excessive market consensus may create risks. The podcast provides a detailed analysis of the $30 billion annual data center service agreement between OpenAI and Oracle, as well as the complex relationship between this transaction and "Project Stargate," revealing the intricate financing, cooperation, and competition among SoftBank, Microsoft, OpenAI, and other stakeholders. The discussion emphasizes that the slow progress of "Project Stargate" stems from the difficult-to-untangle relationships and funding implementation challenges behind it. In addition, the podcast looks forward to the expected release of GPT-5 and its impact on industry expectations, mentions the progress of emerging models such as DeepSeek, and analyzes the challenges and prospects of commercialization in the AI industry given the massive investments, pointing out that funding has become a critical indicator in the current AI competition, and that AI investment in the United States has surpassed the traditional US dollar fund cycle.