Welcome to the DeadCode blog, where I dive into the world of AI, software development, and my personal journey. I write about ML/AI Engineering, and modern web development, along with insights from my journey as a entrepreneur and developer.
Unlike traditional software, AI agents are unpredictable; they might solve the same task in completely different ways each time. This article shows you how to build a proper evaluation framework that tests your agent tools with realistic scenarios, catches both technical and semantic failures, and uses systematic feedback to make your tools actually work reliably.
VS Code's new "Tool Grouping" feature tries to solve MCP tool limits by working around them, but this misses the real problem. Installing multiple MCP servers with dozens of tools doesn't just hit arbitrary limits, it fundamentally degrades LLM performance and inflates costs. Here's why less really is more when it comes to AI coding tools.
AI can feel overwhelming, especially with how fast things move and how often new terms pop up. But itās hard to explore or feel confident around AI if the basic language feels unfamiliar. This list breaks down 50 essential AI terms in simple, clear languageāso you can follow the conversation, ask smarter questions, and feel more at ease navigating the world of artificial intelligence.
I've been using Cursor as my daily coding tool since late 2024, and after plenty of trial and error, I've found four techniques that genuinely boost productivity. From planning a task to parallel task execution, here's what actually works beyond the hype.
After spending weeks testing different AI models for my daily coding and writing tasks, I narrowed it down to two contenders: ChatGPT and Claude. Like I mentioned in my previous article, I'm a firm believer that focusing on mastering one tool is better than juggling multiple options. It lets you really get to know all the features - both current and new ones as they roll out.