RAG is (Not) Dead: How to Think about Building RAG Systems
RAG isn't about vector databases and embeddings, or any specific architecture. It's about retrieving relevant context well.
Kyle's Blog
Musings on engineering, cybersecurity and AI.
Latest posts
Building Kydenticon: A TypeScript Library for GitHub-Style Identicons
Deep dive into creating a zero-dependency TypeScript library for generating beautiful, deterministic identicons with PNG and SVG support, built with modern tooling.
You're Doing it Wrong: Prompt- and Context-Engineer with XML, not JSON
Exploring the syntactic and semantic differences between XML and JSON and why the former provides a more robust structure for complex LLM prompts
MCP Deep Dive: the Great, the Broken, and the Downright Dangerous
I've spent months in the trenches building with MCP. This is my report from the front lines: a deep dive into MCP's powerful vision, frustrating reality, and critical security flaws.
Why Everybody is so Excited about MCP
Everybody's talking about MCP, but what's all the hype for? In this post I dig into what about MCP has everyone so excited.
OpenSearch: Building a Deep Research Agent from Scratch
Building AI Agents is hard! In this post I walk through the process of building OpenSearch, my open-source deep research agent, from scratch. I'll go over system design, my stack, and what I learned about building agents.
I Will Never "Vibe Code" Again: Building a Native MacOS App with Cursor
I decided to "Vibe Code" a port monitor app that lives in my Mac's menu bar, and I learned lots of lessons along the way like "How to debug corrupted system caches" and "How to boot a M3 Macbook Pro into safe mode". Featuring hot takes on vibe coding.
8 Key Components of Secure & Reliable OAuth Integrations
Building OAuth integrations is quite complicated and can be migraine-inducing - specifics vary between implementations, and security is easy to get wrong. In this article, I walk through what you need to make sure you're thinking about.
MCP, A2A, and Inter-agent communication
A podcast with the Naptha AI team about Google's new A2A protocol for inter-agent communication, and how it fits into the agent protocol stack with MCP
Implementing OpenAI-Compatible Tool Calling & Tool Streaming for Open-source models in vLLM
This is a transcription of a talk I gave at vLLM's office hours after landing vLLM's first-of-its-kind tool calling implementation that allows using OpenAI-compatible tools and tool streaming with opens-source models.
A Beginner's Guide to Malware and EDR Evasion
A beginner's guide to bypassing EDR systems using APC queue injection and direct system calls.
Dumping Stored Enteprise Wi-Fi Credentials with Invoke-WifiSquid
Credential re-use is rampant, and every post-exploitation action that gets you credentials for potential lateral movement is a win in my book. Learn how to grab stored enterprise wifi credentials from compromised machines.
Extracting Saved Browser Passwords: Your New Favorite Post-Exploitation Technique
Learn how to extract saved browser passwords from compromised hosts for fun and profit - your new favorite post-exploitation technique!