{"id":24,"date":"2026-04-01T06:00:00","date_gmt":"2026-04-01T06:00:00","guid":{"rendered":"https:\/\/regentis.ai\/blog\/?p=24"},"modified":"2026-09-07T12:43:17","modified_gmt":"2026-09-07T12:43:17","slug":"ai-made-coding-faster-but-is-code-review-moving-at-the-same-speed","status":"publish","type":"post","link":"https:\/\/regentis.ai\/blog\/ai-made-coding-faster-but-is-code-review-moving-at-the-same-speed\/","title":{"rendered":"AI Made Coding Faster, But Is Code Review Moving at the Same Speed?"},"content":{"rendered":"\n
\"\"<\/figure>\n\n\n\n

Software developers around the world are writing code 10 times faster than they were a year ago. Thanks to tools like Copilot, Cursor, and Claude, developers can produce much more output in a short time. From the outside, this naturally looks like faster delivery, faster project progress, and higher productivity. However, there is one point where we cannot keep up with this speed: Code Review.<\/p>\n\n\n\n

Code is being written faster, but in practice project delivery times can stay the same, and sometimes work can even be delayed. For this reason, in this article I want to talk about the fact that while AI is speeding up production, we are not able to move forward at the same speed in parallel.<\/p>\n\n\n\n

Where the code review bottleneck actually moved
<\/h2>\n\n\n\n

According to Faros.ai\u2019s research based on the data of more than 10,000 developers, the use of AI has led to a 98 percent increase in the rate of opening PRs. But delivery time, in other words the speed at which work reaches the customer, is not increasing at the same rate. In fact, in some teams, a decline can even be seen.<\/p>\n\n\n\n

We can actually understand this more easily with the following example: We speed up the production line in a car factory by 10 times. In this way, we make it possible for 1 car to come off the line every second. But we continue with only one engineer at the quality control point. In this case, a pile-up at the quality control point becomes unavoidable. What we are experiencing in the software world today is a bit similar to this. The bottleneck in coding does not disappear, it only changes its place.<\/p>\n\n\n\n

In the same research, while PR volume increased by 98 percent, the time spent reviewing code also increased by 91 percent. The size of a PR, meaning the amount of code that needs to be read and understood in one go, has also grown by 154 percent.<\/p>\n\n\n\n

Looking at this picture, we can conclude that AI has not removed the bottleneck in the system, but simply moved it from the production stage to the review stage.<\/p>\n\n\n\n

Why AI-generated code takes longer to review
<\/h2>\n\n\n\n

When we think that we are getting faster with AI, we often focus only on the moment the code is written. However, software development is not only about producing code. The produced code needs to be read, understood, its risks need to be seen, its fit with the current structure needs to be evaluated, and it needs to be approved safely.<\/p>\n\n\n\n

AI tends to generate a larger amount of code. At times it is overly defensive, at times more verbose than it needs to be, and sometimes it comes up with solutions that are simply bigger than the problem requires. All of this creates additional work for the person doing the review.<\/p>\n\n\n\n

According to CodeRabbit\u2019s research, code written by AI can contain 7 times more errors and 1.5 to 2 times more security vulnerabilities compared to human-written code. Because of this, the review time becomes 4 times longer.<\/p>\n\n\n\n

So if one of my teammates saves 2 hours by coding with AI, but another teammate spends 4 hours understanding, reviewing, and finding that code safe, then the system is not really getting faster. On the contrary, it becomes slower. Something that feels faster at the individual level may not create the same result at the team level.<\/p>\n\n\n\n

The productivity illusion: feeling faster while shipping slower
<\/h2>\n\n\n\n

So why do I still adopt AI so much in my work? Because psychologically, I am falling into an illusion.<\/p>\n\n\n\n

They test this illusion with two teams made up of truly experienced software developers. One team completes the given tasks using only AI, while the other team moves forward with completely human-written code. When delivery time comes, they find that the team using only AI is 19 percent slower. After the tasks are completed, they send a survey to both teams, and in the survey results, the slower team says that AI use made them 20 percent faster.<\/p>\n\n\n\n

This is exactly what falling into the illusion means.<\/p>\n\n\n\n

In these tasks, because the coding is done by AI instead of a human, the time spent on the review process is not seen as extra time in the same way. This is because there is a psychological relief that comes from escaping the boredom of coding and part of debugging.<\/p>\n\n\n\n

There is still much more that could be said about AI. But before I go on, I would really like to hear what you think about this topic.<\/p>\n\n\n\n

What do you think should be the right approach when we cannot move review processes in parallel with the production speed of AI?<\/p>\n\n\n\n

<\/p>\n","protected":false},"excerpt":{"rendered":"

Software developers around the world are writing code 10 times faster than they were a year ago. Thanks to tools like Copilot, Cursor, and Claude, developers can produce much more output in a short time. From the outside, this naturally…<\/p>\n","protected":false},"author":2,"featured_media":0,"comment_status":"closed","ping_status":"closed","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[3],"tags":[27,29,21,25,26,30,28],"class_list":["post-24","post","type-post","status-publish","format-standard","hentry","category-artificial-intelligence","tag-ai-coding-tools","tag-ai-generated-code","tag-artificial-intelligence","tag-code-review","tag-developer-productivity","tag-engineering-bottlenecks","tag-pull-requests"],"_links":{"self":[{"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/posts\/24","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/users\/2"}],"replies":[{"embeddable":true,"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/comments?post=24"}],"version-history":[{"count":3,"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/posts\/24\/revisions"}],"predecessor-version":[{"id":74,"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/posts\/24\/revisions\/74"}],"wp:attachment":[{"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/media?parent=24"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/categories?post=24"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/regentis.ai\/blog\/wp-json\/wp\/v2\/tags?post=24"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}