Guides & Reviews
Practical AI income guides — no hype, just what actually works.
How to Write a Multi-Agent Supervisor: The 39% Rewrite and Why the Savings Weren't Where I Expected
What is a supervisor agent, and when does splitting work into subagents actually pay for itself? I cut my supervisor skill from 3,067 words to 1,872 and found the money was not in reasoning effort. It was in deleting the mandatory root to supervisor to implementer to test-author chain. With the multi agent vs single agent numbers from Anthropic, AWS and a published benchmark that ran the same job both ways.
How to Run Claude Code From Your Phone: Four Official Paths and the Conditions That Break Every One of Them
Can you run Claude Code from your phone? Yes, and Anthropic now ships four ways to do it. Each one refuses to start under a specific condition: an API key instead of a subscription, Bedrock, a proxy in ANTHROPIC_BASE_URL, zero data retention. That gap is where third-party chat bridges like zarazhangrui/lark-coding-agent-bridge and chenhg5/cc-connect still live. What each option requires, and what you hand over when you put a coding agent in a chat window.
Claude Code Skills: What They Are, How to Use Them, and the Minimal Memory Setup We Run
What is Claude Code skills and how to use it? The four-skill memory harness we run in production, file by file: where Claude Code skills are stored, whether Claude uses them automatically, the exact SKILL.md files, and the settings.json hook that makes state survive restarts. Plus two mechanisms people keep configuring that do not exist, and how skills compare to hooks, slash commands, and CLAUDE.md.