DeepSeek V3
MoE model with 671B total parameters
Episodes
-
#6021: Why Chat Templates Break Models SilentlyEvery chat API flattens your messages into special tokens. Get one delimiter wrong and the model just gets quietly dumber. -
#6012: Inside the Decode Loop: How LLMs Generate One Token at a TimeThe model doesn't choose tokens or decide when to stop — the engine does. A step-by-step tour of one decode iteration. -
#5408: When Small NLP Models Beat the LLMFeature extraction, fill-mask, token classification — the classic NLP tasks still have a job. Here's when a small model beats a frontier API. -
#4723: Are Model Families Actually Different Models?Claude Opus, Sonnet, and Haiku aren't trim levels — they're different models sharing a brand name. -
#2380: The Policy Chess Game Behind Your Country's Rainy-Day FundWhat does $15 trillion in global foreign currency reserves mean for fiscal policy and economic stability? We break it down. -
#2379: The Geological Lottery: Why Oil Is So Unevenly DistributedDiscover how oil forms, why it’s concentrated in a few regions, and whether we’ve found it all. -
#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. -
#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. -
#107: The $5.5 Million Breakthrough: DeepSeek’s AI DisruptionDiscover how DeepSeek-V3 is disrupting the AI market with massive cost savings and technical innovations like Multi-Head Latent Attention.