Posts

All the articles I've posted.

• Dungeons, Dragons, and Data: Building an MCP-Powered Knowledge Navigator

  Posted on:March 15, 2025 |  at 11:52 AM

  The Model Context Protocol (MCP) represents a significant shift in how we're approaching AI architecture—moving beyond the endless "my model hallucinates less than yours" debates toward practical systems that acknowledge AI limitations. Unlike RAG, which excels with large static datasets but remains inherently limited to its indexed information, MCP establishes a dynamic two-way communication channel for accessing fresh, real-time data when needed. My weekend experiment hooking Claude to the D&D API demonstrated how this client-server architecture creates clean separation between LLM logic and data sources, allowing developers to add or swap components without rebuilding their entire application. Think of it as HTTP for AI—a standardized protocol that frees us from custom integration work while opening fascinating possibilities beyond mere information retrieval, like controlling creative tools such as Blender or game engines. It's the architectural pattern that separates genuinely useful AI applications from impressive but ultimately limited demos.

• Rails Makes LLMs Boring (And That's Good): A Service Layer Implementation

  Posted on:February 26, 2025 |  at 04:44 PM

  Rails service patterns provide an elegant solution for LLM integration without specialized frameworks or microservices. Our D&D character generator demonstrates how standard Rails conventions can handle complex AI capabilities, proving that "boring" technology delivers powerful results with superior maintainability and development speed.

• Rails Makes Its Wisdom Save: Why ‘Boring’ Technology is Winning the AI Campaign

  Posted on:January 30, 2025 |  at 05:17 PM

  Boring technology might be the future of AI applications. Starting a series building a D&D character generator that challenges our industry's obsession with complexity. Rails - with its predictable conventions, server-rendered HTML, and focus on simplicity - is an excellent match for the generative AI era. Rails' established patterns align well with the needs of generative AI applications. The framework's steady evolution and focus on simplicity offer some interesting advantages for LLM integration. Part 1 examines how Rails' approach to web development might be particularly relevant in the AI era. While our industry often equates innovation with complexity, there's value in examining proven patterns through this new lens.