👋 Mike’s Blog

The Model Context Protocol (MCP) allows large language models (LLMs) to call external tools while answering user prompts. In practice, most people use chat-based tools like Claude or Copilot rather than interacting directly with model APIs, and these tools make it easy to configure MCP servers without writing code. That said, if you want to connect your own data sources or custom functions as MCP tools, or connect your code to MCP tools, you’ll have to roll up your sleeves and do some hands-on work. ...

Based on Gartner Hype Cycle for Cloud Platform Services, 2025 ...

As we integrate services and data APIs into agentic AI solutions, interest is growing in how the Model Context Protocol (MCP) can standardize the way tools expose their capabilities to agents. With that in mind, I’ve assembled—yes, with the help of AI—a survey of key topics and resources related to MCP. MCP is an open standard (launched by Anthropic in Nov 2024) for exposing data sources, tools, and “resources” to AI agents via a uniform interface. It is designed to replace the ad-hoc “one-off connector per tool/agent” pattern, simplifying how LLM-based agents integrate with live systems. [1] ...

Is this just distributed computing re-packaged for AI? What is A2A Without AI Agents? I got into a debate with Gemini recently about Agent-To-Agent protocol (A2A). I said I thought it was a retread of existing distributed computing technologies like Service Discovery, Mesh, CORBA, etc. Perhaps Gemini took it personally, as Google (Gemini’s Creator) had announced A2A in April, and Gemini got a little “gushy” on how it was “a revolutionary new idea.” Also, perhaps “debate” is too strong a word. And I might want to consider getting out more often. ...

Combining deterministic NLP with LLM-as-Judge for robust evaluation Summary Evaluation Challenges Summary evaluation metrics sometimes fall short in capturing the qualities most relevant to assessing summary quality. Traditional machine learning for natural language processing (NLP) has covered a lot of ground in this area. Widely used measures such as ROUGE focus on surface-level token overlap and n-gram matches. While effective for evaluating lexical similarity, these approaches offer limited insight into aspects such as factual accuracy or semantic completeness [1]. ...

In this segment, I’ll generate many candidate applications using my experimental framework, CodeAgents, choosing from a set of models: GPT-4.1, Claude 3.7, and GPT-4o. Then, I’ll compare and contrast the solutions. Along the way, I’ll present some ideas and tips on improving AI-generated code in ways that generally translate to other tools and frameworks. It isn’t easy to score how good an AI-coded solution is. Of the possible metrics, code complexity might not be as meaningful as long as the AI understands the code, as would “maintainability,” as that’s based on human limitations; the AI can refactor on the fly. Test coverage is a good metric as it measures how well the AI-generated test suite covers the code. ...