Long-form interviews and conversations with tech, AI, and finance leaders (30min+)
Former Goldman Sachs Chairman Lloyd Blankfein explains why modern risk management is about contingency planning rather than prediction, warns that AI's untestable leverage poses unprecedented financial dangers, and reflects on how his upbringing in Brooklyn public housing shaped a crisis-tested leadership philosophy.
After decades of decline known as the 'great dechurching,' American religiosity has reached an inflection point where secularization has paused, with Gen Z and young men specifically showing renewed interest in faith driven by pandemic isolation, mortality awareness, and a search for tangible community.
Binance founder Changpeng Zhao discusses seeking a presidential pardon for his felony conviction, explains why he operates from the UAE (citing safety and pro-crypto regulation), and warns that the shift to intangible assets (92% of modern wealth) makes entrepreneurs highly mobile, forcing countries to compete on tax and regulatory policy.
Harvey CEO Winston Weinberg reveals the rigorous prioritization systems and psychological frameworks he uses to run his $11 billion legal AI company, including a massive living Google document for daily re-ranking and the discipline to ignore external pressure in favor of fixing real bottlenecks.
Trump's upcoming summit with Xi Jinping marks a historic shift where China enters negotiations from a position of relative strength for the first time, with both sides seeking economic concessions while deeper security tensions remain unchanged.
This seminar outlines a roadmap for autonomous surgical robotics to address critical healthcare labor shortages, proposing a physics-based approach built on four pillars—perception, modeling, planning, and control—that achieves sub-2mm precision through real-time digital twinning rather than relying on data-scarce foundation models.
Inference now dominates AI economics, with OpenAI generating 8.6 trillion tokens daily—exceeding frontier model training compute in under four days. Unlike training, autoregressive inference cannot parallelize across sequences, making it fundamentally memory-bandwidth bound rather than compute bound, with batch sizes under 295 on H100s failing to saturate GPU capacity.
This lecture transitions from theoretical foundations to practical architecture design for diffusion models, explaining how U-Net structures leverage convolutional inductive biases, hierarchical downsampling for global context, and skip connections to preserve local details while maintaining strict dimensional requirements for iterative denoising.
Shrimai Prabhumoye presents advanced LLM pre-training strategies from her work at Nvidia, demonstrating that curriculum learning (two-phase training) and front-loading reasoning data during pre-training create stronger foundations and durable performance gains that cannot be matched by increased compute in later stages.
Nuaman Tazzy from HuggingFace explains how to scale transformer training to thousands of GPUs using data parallelism strategies, from basic Distributed Data Parallel (DDP) to Fully Sharded Data Parallel (FSDP/ZeRO), emphasizing memory optimization techniques and the critical importance of overlapping communication with computation to keep GPUs fully utilized.
Bill Fleckenstein warns that algorithmic passive investing flows and QE have created a 'giant mindless robot' driving stocks to all-time highs regardless of war risks or fundamentals, but predicts catastrophic consequences when the bond market finally revolts against unsustainable US debt levels.
Ben Horowitz explains how Andreessen Horowitz rearchitected venture capital by treating entrepreneurs as customers, centralizing control to enable scaling, and bootstrapping network effects through asymmetric early investments.