Cloud-based coding agents from Claude Code on the Web, Cursor Cloud Agents, and Codex all promise the same thing: kick off a task, walk away, and come back to a completed pull request. But do they actually deliver? And more importantly, which one fits into your real-world development workflow?
Claude just launched Skills across all their platforms, and I've spent a week testing the Claude Code implementation specifically. Do Claude Code Skills actually solve problems developers have, or are they better suited for consumer platforms like Claude.ai? I compare Skills to the developer tools we already use...
I demonstrate my workflow for running multiple AI coding agents (Claude Sonnet 4.5 and Cursor's Composer 1) in parallel on the same feature, then comparing their work to choose the best implementation. The focus is on solving the practical challenges of reviewing agent work in git worktrees by using a branch-based approach that makes testing, debugging, and merging much more manageable.
In this video, I share a best practice for managing and customizing your standards in AgentOS, which simplifies the process of updating to the latest version (it was 2.0.5 at the time of this recording). I demonstrate how I commit my base installation to a private GitHub repo for easy...
Claude Code is excellent as an autonomous coding agent. So the obvious next step is to run multiple Claude Code agents in paralell. Simple, right? This video shows you how to use Git Worktrees and my method for automating the workflow for spinning up new worktrees to run multiple Claude Code agents simultaneously.
Training & community for pro software developers building with AI.
Created by Brian Casel (that's me). I'm a career software developer, founder, and creator of Builder Methods and Agent OS, the system for spec-driven development with AI.
