code verification (CR)

in English: code review

basicwidespread

Code review is the step where a completed code segment is examined by someone other than the author—a person or AI—before being merged into the shared codebase.

In plain terms

Think of the moment you show an important email to a colleague before sending it: they notice the typo, the ambiguous sentence, or that it is going to the wrong recipient. Code review is the same in programming: a second pair of eyes reads the written code before it enters the shared system.

The goal is not just debugging. The reviewer also checks whether the code is understandable, secure, fits with other parts, and whether someone else will be able to navigate it in a year. Meanwhile, knowledge spreads: more people learn how things work.

With the emergence of AI, the situation has reversed. Previously, a human wrote the code and a human reviewed it; now increasingly a coding agent writes it, and the human's job is the review – or another agent does the first round, and the human only decides on the sensitive points. Since the machine writes much faster than a human can read, verification has become the bottleneck.

How it works

The author (human or coding agent) completes a modification and submits it for approval instead of committing directly to the shared codebase; this is often called a pull request.
The reviewer sees what changed: the old and new lines side by side, with a brief explanation from the author about what was done and why.
Meanwhile, automated checks also run: automatic tests, style and security reviews, and increasingly, a first-pass review by an AI agent.
The reviewer writes comments next to individual lines: asks questions, flags errors, suggests better solutions. The author makes corrections, and this repeats until there are no open questions.
If everything is in order, the reviewer approves the change, and it is merged into the shared codebase; the decision leaves a trace and can be retrieved later.

An everyday example

After updating your banking app, you see a new button for transfers. Before this change reached your phone, someone else reviewed the underlying code and approved it – hopefully catching if the amount was destined for the wrong account.

Who is affected

developerAn increasing part of your daily work is not writing code, but reviewing and approving code written by agents. → Code

entrepreneur, executiveIf your company develops or orders software, the quality of verification determines whether errors in quickly written code are discovered by you or your clients.

jobseeker, new graduateToday, even junior programmers are expected to critically read code written by others and by machines, not just write it. → Work & HR

everyoneThe reliability and security of the applications you use partly depend on whether someone reviewed their code before release.

A common misconception

MythIf AI wrote the code and another AI reviewed it, a human no longer needs to look at it.

In realityAn AI reviewer catches many errors but also gets things wrong, and often does not know what the client actually wanted. Responsibility stays with whoever approves the change.

Limits and risks

Oversight is not a guarantee: careful review can also miss errors, especially with major changes.
If there is too much code and too little time, the review can easily become a mere formal nod – a real risk given the volume of code produced by AI.
An AI reviewer can confidently flag errors that do not exist and miss real ones, so its comments, too, must be weighed.
Reading does not replace testing: code that looks good may not necessarily work.

Similar, but not the same

testing – Testing runs the program and checks if it works correctly; code review reads the written text itself.
static code analysis – This searches for known bug patterns mechanically according to prewritten rules, while code review requires judgment on whether the solution is good.
security audit – An audit is a less frequent, comprehensive examination specifically seeking vulnerabilities, whereas code review is a daily check of every single modification.

Related concepts

Broader

software development, quality assurance, human oversight

Narrower

AI code review, security code review

On the edge

pull request, coding agent, pair programming, testing, version control, static code analysis

A brief history

Michael Fagan described the method of formal code review at IBM in 1976. The modern, lighter online version spread through open-source projects and with pull requests after GitHub launched in 2008.

Where things stand

Human code review has been standard practice for decades, and by the end of 2026, the first-round review by AI agents has also become common in many development teams. The main question today is how humans can keep up with the volume of code written by agents.

Status: 2026 November

In the news

Appears in 33 Agent Age stories; the curve shows weekly occurrence in the news magazine.

GitHub Copilot impact dashboard adds chat and VS Code agent usage, code review gets API controlsNov 6, 2026 · Code
AI code review is the next bottleneck: more code, fewer lines readNov 2, 2026 · Code
Qodo builds context into AI code review with CodexNov 2, 2026 · Code
Atlassian: AI time savings are disappearing into code reviewOct 28, 2026 · Code
Greptile: DIY AI code reviewer costs $3 to $25 per pull request and still misses the key bugOct 26, 2026 · Code
AI code review finds bugs, delaying PostgreSQL 19Oct 19, 2026 · Code
Anthropic makes Code Review in Claude Code generally availableOct 15, 2026 · Code
Copilot administration: model and MCP controls, server-side settings, organization billingOct 12, 2026 · Code

Also known as

code verification · code review · code review · code review · code review · peer code review · pull request review · CR

Sources

Google Engineering Practices: Code Review Developer Guide

This entry was machine-edited and has undergone machine expert review; the “In the news” section updates automatically from Agent Age stories.