How to understand code your AI coding agent wrote
Read one run at a time, right after it finishes, starting from what you asked for. For each important change, write one sentence about what a user can now do. Read the risky files first. BuildGist writes that first read for each meaningful run.
One run, one sentence
Start from the request
Reread what you asked for before any of the code. The question is whether the change matches that request, and what it did past it.
List what changed
git status
git diff --stat
That is the files and the size of each change. Committed runs are covered in what did Claude Code change?
Read in order of risk
Read sign-in and permissions first, then anything that deletes data, then payments, new dependencies and configuration. Read interface text and docs last.
One sentence for each important change
Finish "Now a user can ..." for each important change. If you cannot, that is the part to ask about, naming that change rather than asking for the whole run again.
On the BuildGist home page, a real Claude Code run, shortened and edited, asked to "Add login with Google." 7 files changed, +224 -16. Two sentences cover what that example shows:
- Now a visitor can sign in with Google, and the app never sees their password. Google checks who visitors are and vouches for them. The name for that is the OpenID Connect (OIDC) authorization code flow.
- Now any signed-in user can delete anyone else's note, because nothing checks that the note is yours.
The second sentence is the high concern in that example. Sign-in working does not show it.
Keep up with each run
One read when the run ends stays short. Leaving several runs unread means the next read is all of them at once.
BuildGist writes the first read
After each meaningful run of Claude Code, Cursor or Codex, BuildGist writes a plain-English explanation of what changed, what it means and what to check, and puts it in your Inbox. It works from uncommitted work too, and small changes stay quiet. It is written by AI, so check important behavior against the code and tests.
What you need
- A Mac with Apple silicon and macOS 12 or later.
- Claude Code, Cursor or Codex.
- A Git project. No commit, push, remote or GitHub repository is needed.
- A GitHub account, to sign in. BuildGist does not access your GitHub repositories.
Codex needs you to approve BuildGist once, in Codex. BuildGist shows the steps. Setup is in Getting started.
To explain a run, your computer sends BuildGist what that run needs, and BuildGist passes it to Anthropic. BuildGist does not keep this material. It does not send your whole repository, files the run did not change, or files named like secrets, such as .env files and keys. It keeps the explanation and short cited excerpts of changed code, at most 12 lines each. Secret masking is best-effort. Privacy has the full list.
Every account gets 7 days of BuildGist Pro free, with no card. The 7 days start at your first explanation or Continue handoff. Nobody is charged when the trial ends. After that, BuildGist Pro is $12 per month, plus any applicable tax shown at checkout. See pricing.
Apple silicon, macOS 12 or later. United States, 18 or older.
Related
Questions: kayvanandre@gmail.com.