mamba
State-space model architecture
Episodes
-
#2672: When a Startup Claims to Break the Quadratic WallA startup claims linear attention scaling at 12M tokens, beating GPT-5.5 on retrieval benchmarks. -
#2366: Why LLMs Forget the Middle of Long ConversationsWhy do large language models struggle with the middle of long conversations? Explore the science behind attention dilution and practical fixes. -
#2350: NVIDIA's Strategic Pivot: From Chipmaker to Model BuilderDive into NVIDIA’s Nemotron 3 Super, a hybrid MoE model combining Mamba, Transformers, and multi-token prediction for cutting-edge efficiency. -
#2065: Why Run One AI When You Can Run Two?Speculative decoding makes LLMs 2-3x faster with zero quality loss by using a small draft model to guess tokens that a large model verifies in para... -
#2063: That $500M Chatbot Is Just a Base ModelThat polite chatbot? It started as a raw, chaotic autocomplete engine costing half a billion dollars to build. -
#2061: The Memory Bottleneck That Drives Attention DesignAttention is the engine of modern AI, but it’s also a memory hog. Here’s how MQA, GQA, and MLA evolved to fix it. -
#1799: The Original AI Blueprints: BERT & CLIPBefore GPT, two models changed everything. Discover how BERT and CLIP taught machines to read and see the world. -
#1604: The $3 Billion Stealth Giant: AI21 Labs & NvidiaWhy is Nvidia eyeing a $3B deal for AI21 Labs? Discover the tech behind the "OpenAI of Israel" and their revolutionary hybrid architecture. -
#1547: Why AI Stopped Reading and Started Seeing EverythingFrom sequential bottlenecks to parallel powerhouses, discover how the Transformer architecture revolutionized how machines process the world. -
#1111: Surviving the arXiv Deluge: Finding Signal in AI's Paper FirehoseDiscover the unsung research papers that built the AI era and learn how to navigate the relentless flood of new machine learning breakthroughs. -
#1088: Why AI Can Read a Library but Only Write a PostcardDiscover why frontier AI models can process millions of words but struggle to write more than a few pages without losing their logical thread. -
#1081: The K-V Cache: Solving AI’s Invisible Memory TaxWhy does your AI get slower as you chat? Discover the K-V cache, the invisible bottleneck of generative AI, and how we're fixing it in 2026. -
#1080: From Prompt to Intent: The Claude Opus RoadmapExplore the engineering roadmap from Claude 4.6 to 5.0 as AI evolves from a simple chatbot into a fully autonomous cognitive partner. -
#808: The AI Deprecation Trap: Anthropic vs. GoogleIs your AI model about to retire? Explore how Anthropic and Google handle model sunsets and what it means for your production code. -
#608: When AI Eats the World's MemoryWhy is a 32GB RAM kit now $400? Herman and Corn dive into how OpenAI is gobbling up 40% of the world's memory supply for its "Stargate" project. -
#182: The Quadratic Bottleneck: Why AI Needs New ArchitecturesIs the transformer’s reign ending? Herman and Corn explore Mamba, x-LSTM, and the architectures solving AI's massive memory problem. -
#126: The Cocktail Party Problem: Why AI ForgetsWhy do AI models "lose the plot" after a few thousand words? Discover the mechanics of attention and the innovations solving context window limits. -
#122: The Comeback of RNNs: Why Old Tech Still MattersHerman and Corn pull back the curtain on AI to explain the mathematical "plumbing" of neural networks and the future of machine learning. -
#111: Beyond Transformers: Solving the AI Memory CrisisWhy does AI forget your conversation every time you hit enter? Herman and Corn dive into the "stateless" nature of LLMs and the future of memory.