This episode of Unsupervised Learning interviews David Luan, head of the AGI SF Lab at Amazon, about the implications of DeepSeek, the current state of AI agents, and the future of AGI. The conversation covers the initial market reaction to DeepSeek, Luan's perspective on combining LLMs with reinforcement learning to create more capable and reliable agents, and the challenges in building a robust "agent factory." Luan emphasizes the importance of reliable agents for practical applications and predicts that while AGI is not far off, its widespread adoption will be slower due to societal and process limitations. A key takeaway is Luan's belief that focusing on building a reliable "agent factory" and addressing the engineering challenges of agent development are crucial for progress.