AI Jason

YouTube creator focused on LLM application architecture — specifically the design patterns that sit between raw model APIs and production AI products. Strong opinions on RAG vs alternatives, MCP server design, and cost-aware LLM ops. Runs the AI Builder Cloud community.

Channels

  • YouTube: AI Jason — LLM app architecture, CAG, MCP patterns, prompt engineering
  • Community: AI Builder Cloud

Content in This Wiki

Key ideas

  • CAG over RAG: when the dataset fits the context window, pre-loading the entire corpus is cheaper, faster, and more accurate than chunked retrieval
  • MCP as the wrapper: data-access patterns (CAG, RAG, hybrid) all benefit from being exposed as MCP servers so any compliant client can use them
  • Cost-economics framing: long-context model pricing collapses (Gemini 2.0 Flash at $0.01/M) are what flipped the RAG-vs-CAG calculus

See Also