Guides
Working with your agent, without losing the plot.
Agents write code faster than anyone can follow. These guides are about the other half: reviewing it, understanding it, and keeping the decisions yours.
AI code reviewHow to review code your agent wroteReviewing AI-generated code is not like reviewing a colleague's. Here is what to look for, why diffs and summaries fall short, and how to see the decisions your agent made.Understanding AI codeUnderstanding code your agent wroteWhen an AI agent writes most of the code, understanding it becomes the hard part. A practical way to rebuild your mental model of what changed and why.Large pull requestsHow to review a 20,000-line pull requestHuge PRs get rubber-stamped. A practical approach to reviewing very large pull requests by concern, in chapters, with the decisions shown on the lines.Claude CodeReviewing what Claude Code changedClaude Code can change dozens of files in one session. How to review its work properly: see each decision on the lines, push back where you disagree, and keep the judgement yours.Codex, Cursor, AmpReviewing what Codex, Cursor or Amp changedWhichever coding agent you use, the review problem is the same: fast changes, long summaries, decisions you never made. How to see them on the lines instead.LGTM fatigueHow to stop rubber-stamping pull requests"Looks good to me" has become the review. Why rubber-stamping happens when agents write the code, and a practical way to review decisions instead of approving them.AI code qualityKeeping code quality when your agent writes most of itAI agents write working code fast. Quality slips in the decisions nobody made on purpose. How to keep judgement over a codebase an agent mostly writes.Vibe codingVibe coding without losing track of your systemVibe coding is fast and fun until you cannot explain your own code. How to keep the speed and still know what you shipped.OnboardingLearning a codebase you didn't buildNew team, unfamiliar service, the person who built it has left. How to get your agent to walk you through one real request end to end, on the actual code.DebuggingDebugging with your agent: see the root causeYour agent found the bug and wrote five paragraphs about it. How to get the root cause shown on the lines, with the timeline of what happened, so you can confirm it.PlanningPlan with your agent before it writes codeThe cheapest time to disagree with your agent is before the code exists. How to review a plan as code plus consequences, and decide with both numbers in front of you.CompatibilityWorks with the agent you already usedeck works with any coding agent that can run a shell command, including Claude Code, Codex, Cursor and Amp. No new model, no API key, no account.CodeRabbit alternativeA different kind of CodeRabbit alternativeCodeRabbit and similar bots comment on your pull request after the fact. deck is the step before: your own agent shows you each decision on the lines, and you decide. Free and open source.AI code review toolsAI code review tools, compared honestlyCodeRabbit, Greptile, GitHub Copilot code review, Graphite, Qodo, Cursor BugBot and deck do different jobs. What each kind of tool is for, and how they fit together.Review fatigueWhen AI code review becomes noiseReview bots that comment on everything train people to ignore them. Why AI code review gets noisy, and how to get signal: fewer claims, shown on the code, that you decide on.Comprehension debtComprehension debt, and how to pay it downComprehension debt (or cognitive debt) is the gap between the code your agent wrote and what your team understands. Why it grows, why it bites, and how to pay it down as you go.The review bottleneckReview is the bottleneck nowAgents write code faster than anyone can read it, so the pressure moves to review. Why more code means less understanding, and how to review at the speed agents write.