Key Moments
Behind Latent Space, ft AIEi Shanghai preview — with swyx and Galen Gao
Latent Space PodcastKey Moments
AI agents can now replicate software tools, threatening the SaaS industry by enabling rapid, low-cost cloning. This shifts the business model from selling software to becoming an AI agent developer or specialized data provider.
Key Insights
The AI Engineer (AIE) conference is expanding globally, with Shanghai hosting its first event, aiming to bridge the gap between AI advancements in China and the global community.
The concept of an 'AI Engineer' emerged from recognizing the growing need for individuals who can harness AI models, distinguishing them from non-technical roles like prompt engineers.
Latent Space began as a blog for openly learning about AI, evolving into a podcast and community that emphasizes sharing knowledge and mistakes, contrasting with the secrecy often found in finance.
The 'Kill My SaaS' initiative demonstrated that AI agents can replicate existing software tools, with 69 submissions showcasing the potential to disrupt the SaaS market by offering customized, low-cost alternatives.
The development of 'System 1' AI models, as championed by Jeff from Typesafe, represents a shift towards integrating AI directly into code logic, making software smarter overall rather than relying on a single conversational interface.
The AI industry faces potential 'winters' similar to past tech cycles, and broader concerns about AI safety and alignment are paramount, with some viewing advanced AI as a potential existential risk due to its rapid, exponential development.
The rise of AI Engineer conferences and global collaboration
The conversation opens by highlighting the significance of the AI Engineer (AIE) conference in Shanghai, marking its debut in China. This event underscores a desire to foster greater dialogue and knowledge sharing between China's burgeoning AI sector and the global community. Shawn (swyx) expresses gratitude to local organizers like Standard Partners and acknowledges the work of individuals like Ming and Liu, who were instrumental in bringing AIE to Shanghai. The initiative aims to showcase China's progress not only in open-source models but also in robotics and other startup ventures, while also recognizing the unique entrepreneurial culture and demanding work ethic, often referred to as '996,' that characterizes some Chinese tech companies.
From public learning to community building: The Latent Space origin story
Shawn shares the genesis of Latent Space, emphasizing the principle of 'learning in public.' This approach, he contrasts with the zero-sum nature of finance, posits that sharing knowledge in technology increases its value. His early experiences, including a less-than-ideal management situation at Two Sigma, fueled his drive to write and speak openly about his learning process. This led to the creation of Latent Space, initially as a blog. The pivotal moment for its growth came with the emergence of Stable Diffusion, predating ChatGPT. By analyzing and documenting his learning journey on the blog, Shawn gained recognition, notably from Stability AI's CEO, Emad Mostaque, leading to the podcast's eventual formation and the broader Latent Space community. This philosophy of open learning is particularly contrasted with Asian cultures that often emphasize formal qualifications.
The emergence and evolution of the AI Engineer role
The concept of an 'AI Engineer' is explored as a response to the increasing demand for individuals capable of harnessing AI models. This role is positioned as distinct from prompt engineers, emphasizing the ability to orchestrate and program multiple AI models simultaneously. Shawn traces the term's origin to a conversation with a non-technical woman interested in AI, realizing the significant advantage engineers would have. The discussion highlights how this title provides a clear career path and identity for those aspiring to work with AI, differentiating from more research-focused roles. The AI Engineer Expo has grown significantly, reflecting the demand for this specialized skill set, with numerous tracks covering areas like agent programming, RAG, and multimodal systems, indicating the evolving nature of the field.
Disrupting SaaS: The 'Kill My SaaS' challenge and its implications
A significant portion of the discussion revolves around the 'Kill My SaaS' competition, which aimed to demonstrate the power of AI agents in replicating existing software tools. Shawn, as a business owner, highlights the substantial costs associated with traditional SaaS subscriptions, noting that 10% of his conference budget—around $300,000 per event—is spent on software. The competition revealed that AI agents could clone these tools for a fraction of the cost ($10,000 in one instance) and be customized using agent programming. This poses a serious threat to the SaaS industry, particularly for small and medium-sized businesses (SMBs). The implication is a fundamental shift where businesses may choose to build or clone their own tools rather than relying on expensive, off-the-shelf solutions, or SaaS companies will need to pivot to become AI agent developers themselves.
The shift to 'System 1' AI and the future of intelligent software
The conversation delves into the 'System 1' AI model, pioneered by Jeff from Typesafe. This approach re-envisions AI not as a standalone tool to be queried, but as an integrated component within the core logic of software. Unlike traditional SaaS models that often feature a chat interface bolted onto existing systems, 'System 1' models are embedded directly into the codebase, making the entire software smarter. This contrasts with the 'big model' paradigm, advocating for smaller, distributed models that enhance overall functionality. The reliability and pervasive intelligence offered by this model are presented as more crucial than raw speed, suggesting a future where software is inherently more intelligent at every level, rather than just having a single smart interface.
The power of taste and curated innovation in AI
The concept of 'taste' in product development is explored, defined as the ability to efficiently narrow down vast possibilities to a clear, optimal solution. This is exemplified by the iPhone's transition to a virtual keyboard, a decision that required significant foresight and a rejection of existing paradigms. The discussion highlights how companies like Cursor, by integrating VS Code and developing features like 'Crockpot,' demonstrate this high taste, leading others to imitate their innovations. Similarly, Lagora's focus on streamlining the tedious document review process in the legal field showcases an understanding of user pain points and a deliberate choice to tackle overlooked problems, reinforcing the idea that true innovation lies in identifying and executing the right solutions among many possibilities.
Navigating the complexities of global AI development and cultural exchange
The conversation touches upon the unique dynamics of AI development in China, noting differences in preferred tools (e.g., Comfy UI vs. Canva) and the sheer scale of communities. The intense work culture in China, often characterized by '996,' is contrasted with the U.S. The potential for existential risk from rapidly advancing AI is also raised, with Shawn emphasizing the historical pattern of less advanced civilizations being superseded by more advanced ones. The importance of open dialogue and cultural exchange in AI is stressed, particularly in bridging the gap between Eastern and Western perspectives, despite geopolitical complexities. The desire for more discussions on foundational AI infrastructure, safety, and payment systems, especially from an Asian context, is expressed.
Mentioned in This Episode
●Products
●Software & Apps
●Companies
●Organizations
●Books
●Concepts
●People Referenced
Common Questions
AI Shanghai is the first AI Engineering event in China, presented by Standard Partners. It aims to foster dialogue and share advancements between the East and West.
Topics
Mentioned in this video
More from Latent Space
View all 273 summaries
35 min⚡️ The 2026 Software Factory Infra Stack — Zach Lloyd, Warp.dev
68 minBuilding Legal Agents When You Can't See the Data — Jacob Lauritzen, CTO of Legora
27 minUsing Agents for Long-Horizon Financial Work — Aman Singh, Arca
28 minBuilding ChatGPT for Finance — Joseph Kim, OpenAI's Head of Financial Services
Ask anything from this episode.
Save it, chat with it, and connect it to Claude or ChatGPT. Get cited answers from the actual content — and build your own knowledge base of every podcast and video you care about.
Get Started Free