Why I built Hyperpowers
Most of the code on my team is now written with Claude Code, and reviewing it became the slow part. Why I built Hyperpowers, what it does and what it doesn't.
Blog
How to read a pull request well, and what to check in code an AI wrote.
Start here
A practical method for reviewing a pull request written by Claude Code, Copilot or another agent: get a summary, read in run order, check the why, then decide.
Most of the code on my team is now written with Claude Code, and reviewing it became the slow part. Why I built Hyperpowers, what it does and what it doesn't.
The order I work in on a pull request too big to read top to bottom: what to do before the diff, how to sort the files, what to read closely and what to skip.
Bot reviewers find problems on single lines. The person who approves a pull request needs to know what it does. Why comments can't tell them, and what helps.
What a review with Claude costs in Hyperpowers, measured on a small sample pull request, what makes it cost more, and how to keep the cost down.
Bot reviewers like CodeRabbit and Copilot, reading the diff on GitHub, and guided review with Hyperpowers: what each is for, and when a person should review.
Review a 40 or 70 file GitHub pull request in VS Code with Hyperpowers: paste the link, read it change by change in run order, and submit one review.
An AI agent changed 40 files and nobody remembers why. How Hyperpowers finds the Claude Code request behind each changed file, and what it leaves out.
AI coding assistants sometimes call functions and options that don't exist in your version of a library. Checks a reviewer can run to catch them before merge.
What to check when you review code from Claude Code, Copilot or another assistant: changes nobody asked for, made-up APIs, empty tests and swallowed errors.
A step-by-step way to review a big pull request without skimming: where to start, what order to read the files in, what to look for, and how to comment.
Follow new posts with the RSS feed.