In this episode, we introduce vLLM, an open-source library designed to dramatically improve the speed and efficiency of large language model (LLM) inference. We break down how vLLM uses techniques like PagedAttention to optimize memory usage, increase throughput, and reduce latency—making it ideal for serving LLMs in production environments. Whether you're building AI-powered applications or scaling agentic systems, this episode explains why vLLM is becoming a go-to solution for cost-effective, high-performance model deployment.

In this episode, we explore how 3D AI chips are overcoming the long-standing “memory wall” that limits AI performance. By stacking memory and compute vertically, these next-generation architectures dramatically reduce latency, boost bandwidth, and improve energy efficiency. We break down why traditional chip designs struggle to keep up with modern AI workloads and how 3D integration is unlocking faster training, real-time inference, and scalable agentic systems. Tune in to understand why breakthroughs in hardware, not just software are shaping the next era of AI innovation.

In this episode, we explore Andrej Karpathy’s evolving vision for software development—from “vibe coding,” where developers collaborate intuitively with AI, to full-scale agentic engineering, where AI agents take on autonomous building, debugging, and optimization tasks. We break down how this shift changes the role of developers—from writing every line of code to orchestrating intelligent systems. Tune in to understand how AI-native workflows are redefining productivity, creativity, and the very craft of software engineering.