Blog
Essays and build notes. The product shows up where it earns a mention, and not before.
AI coding policy for teams, written short
An AI coding policy for teams is a short page: which tools, what never goes in a prompt, a named owner, and a board that shows the work. A 30-page doc will not get read.
Claude Code to Jira won't fix your process
Claude Code to Jira reads the ticket faster. The sprint, the review queue, and a hand-updated board stay built for humans typing the code.
Replace the daily standup with what shipped
Replace daily standup status talk by reading what shipped, what is on it, and what is stuck. Meet only on the stuck items and the decisions a person still has to make.
A living project plan beats a perfect one
A living project plan is only useful while it is still true. Keep it rough on the board, review the card before an AI agent builds, and let GitHub move the status.
A markdown backlog for AI coding agents
A markdown backlog for AI coding agents works for one person. Shared decisions and missed intake break it. Keep the rules file short and put each decision on the card.
Project manager accountability when you don't do the work
Project manager accountability is a name on the card. You answer for the date. Someone else does the work. Stuck work comes back to that person.
What coding agents should decide, and what they should ask you
Human in the loop AI coding agents either get a novel to rubber-stamp or redo a call the team already made. Write down what the agent decides, keep each ask to one decision, and leave the merge with a named person.
Stop counting pull requests now that agents open them
Pull requests as a performance metric measure nothing once agents open them. Count cards a person asked for that the team merged, then check the work held.
A status update template for agents: BLUF, then SBAR
An AI agent status update should be skimmable. This status update template puts the bottom line first, then an SBAR handover, and names the next decision.
Name a human owner before the agents open pull requests
Agents write faster than teams review. Put a person on the card before the run starts, keep the merge with that person, and leave AI-on-AI review for comments, not the gate.
Conductor is winding down. Where to run your parallel coding agents next
A Conductor alternative is the queue, not another local runner. Run multiple Claude Code agents in parallel in their own worktrees, and keep who is waiting and which pull request to review on one board.
Let Claude Code and Codex pick up tickets from your board with MCP
A project management MCP server lets Claude Code, Codex, and Cursor pull a card from the board. Here is Kahoona's setup, the tools, and how a named pull request moves that card. A person merges.
AI code review vs verification: check the card, then the diff
AI code review vs verification is two jobs. Abstract agent review produces nits and drifts off the card. Check the intent, the run, and the evidence, then let a person merge.
Sprint planning for small teams when agents ship in hours
Sprint planning for small teams stops fitting once an agent finishes a card in hours. Keep a live board, a cap on review, and the decision that is blocking the work.
How to manage multiple Claude Code sessions without watching them
Manage multiple Claude Code sessions without babysitting the spinners. Parallel AI coding agents leave questions and pull requests on one board, and blockers on the standup.
A Claude Code kanban board: who owns the next task?
A Claude Code kanban board only helps if you can see who owns the next task. Assign the card, let one agent claim it, and don't call it done without a pull request and green checks.
How to review AI agent pull requests without drowning
How to review AI agent pull requests when Claude Code, Codex or Cursor open more PRs than your team can read. Ask for proof, sort PRs by risk and cap what agents get.