AI, Cloud & Code
Deep technical articles on AI agents, cloud architecture, and developer tools. Written by zhuermu.
Categories
Latest Posts
View all →The Real Showdown After the Source Leak: Claude Code vs OpenClaw, 510K Lines vs 530K Lines
In our [previous article](https://zmead.com/blog/openclaw-vs-claude-code-design-philosophy), we dissected OpenClaw's 450K lines of open-source code alongside the 183 Markdown config files in Claude Code's public repository. We knew at the time that Claude Code's core runtime was closed-source — all we could analyze was the plugin ecosystem's definition layer. On March 31, 2026, security researcher Chaofan Shou discovered a 60MB `cli.js.map` file buried inside Claude Code v2.1.88's npm package — 1,884 TypeScript source files, 512,664 lines of code, fully exposed. This time, we finally get to crack open Claude Code's real engine and do a truly apples-to-apples source code comparison.
Complete Failed Request Logging and Async Replay with CloudFront and Lambda@Edge
A dual Lambda@Edge architecture for recording full request headers and body of failed requests — WAF blocks and origin errors — without modifying origin code, with async replay from S3.
WeChat Meets OpenClaw: What One CLI Command Tells Us About Platform Strategy in the AI Agent Era
WeChat's native integration with OpenClaw signals a major shift in platform strategy. Here's why the world's largest messaging app opening up to AI agents matters.
5 Pitfalls of Logging Failed Requests with CloudFront + Lambda@Edge
How we built a dual Lambda@Edge solution to capture full request headers and body for failed CloudFront requests — and the 5 pitfalls we hit along the way.
How AI Coding Agents Actually Work: A Source Code Deep Dive
From LLM reasoning and tool calling to context window management and security models — a comprehensive look at the internals of AI coding agents, based on Amazon Q CLI and Claude Code source code.
Building an Enterprise Agentic AI Platform with Kiro and AWS
How we built a delivery-focused AI agent platform using Kiro IDE, Strands Agents, Amazon Bedrock AgentCore, and AWS CDK — in one week, with zero hand-written code.