Articles
Notes on building with agents.
Practical guides and a few opinions on running Claude Code and Codex in parallel, keeping your repository clean, and spending your attention only where it's needed.
How to run multiple Claude Code agents in parallel
Worktrees, ports and databases, merging one branch at a time, and the real limit: knowing which session needs you.
5 min read
Git worktrees for AI coding agents
The commands, the gotchas, a safe merge loop, and how to clean up without losing work.
4 min read Essay · Kanban for AI agentsWhy ClankDesk is not a Kanban board
Columns are for people moving work. Agents need a board ordered by attention.
4 min read Essay · Managing sessionsToo many Claude Code tabs
A dozen tabs, each with a little title. Quick fixes, and the deeper one.
4 min read Guide · Code reviewReview AI-generated code before it merges
Agent finished is not the same as you approved. A review routine that scales.
4 min read Guide · Two agents, one repoClaude Code and Codex, together
Shared instructions, separate worktrees, two safety models, and clean handovers.
4 min read Guide · Agent safetySafe permission prompts for coding agents
What to allow on its own, what should always ask, and why skipping permissions rarely pays.
5 min read Guide · Voice codingVoice coding: speak tasks into existence
How to brief agents out loud, what not to dictate, and why talking makes better prompts.
5 min read Guide · Vibe codingA vibe coding workflow that scales
Keep the vibe for the asking. Add structure to everything after.
4 min read Explainer · Agent orchestrationWhat is an AI development command center?
The layer above your coding agents, what it must do, and how to evaluate one.
4 min read