MCPs are Boring - Why We Are Losing the Sparkle of LLMs

Explore how mainstream adoption of Large Language Models in coding and agent development is constraining their true potential in this 29-minute conference talk. Discover why current approaches to MCPs (Model Context Protocols) and AI agents are creating limiting "cargo culted boxes" that diminish the transformative capabilities of LLMs. Learn innovative ideas and techniques to break free from conventional thinking and unlock the full power of these AI systems. Gain insights into harnessing what the speaker calls "the shoggoth" - the raw, untamed potential of LLMs that goes beyond typical implementation patterns. Challenge your assumptions about AI development and discover new approaches to working with language models that preserve their inherent creativity and problem-solving capabilities rather than boxing them into predictable, mundane applications.