
Six Things I Argued About While Learning Claude Code
Every Claude Code guide gives the same advice. Here is where I landed after pushing back on each piece of it, and the one idea underneath all six.

Every Claude Code guide gives the same advice. Here is where I landed after pushing back on each piece of it, and the one idea underneath all six.

A hands-on guide to pointing Claude Code at free-tier models through OmniRoute: auto-routing with fallbacks, stacked RTK→Caveman compression, and how to verify all of it actually works.

A practical, step-by-step guide to spec-driven development with GitHub Spec Kit and Claude Code: install it, run the constitution, specify, clarify, plan, tasks, analyze, and implement commands, know what to review at each step, and keep the spec alive as new requirements and bugs arrive.

A practical guide to how memory works in Claude: working, semantic, procedural, and episodic, with pros, cons, token costs, and best practices across Claude Code, Claude Desktop, and Claude Cowork.

portfolio-copilot is a free, open-source Claude plugin for Indian retail investors: daily briefings, stop-loss audits, mutual fund rebalancing, and risk checks. Works live with broker MCPs or with any broker via statement upload.

How I used Claude to audit 642 files across a cluttered Google Drive, detect duplicates, rename ambiguous files, and reorganize everything, plus the one technical blocker that almost stopped it before it started.

Build a custom two-line status line for Claude Code so you can see the model, folder, git status, and session cost at a glance, handy when you're juggling several Windows Terminal tabs. Includes a working Node.js script.

A practical breakdown of GitHub Copilot's two agentic modes (synchronous Agent Mode in your IDE, and the asynchronous cloud Coding Agent) and how to pick the right one.

A practical guide to building MCP servers with .NET Minimal APIs: tools, transports, auth, and testing with MCP Inspector

Arun Endapally, Senior Technical Architect, writing about building software and experimenting with AI