DORA Report: AI Code Review Time Jumps 441%
- The Real Numbers Behind AI Code Review
- Why Review Time Is Exploding, Not Shrinking
- The Benchmark Problem: Nobody Agrees What “Accurate” Means
- What This Means If You Run an Engineering Org
- The Contrarian Case: 19% Slower, Not 20% Faster
- GitHub Code Quality’s July Launch: A Real Test Case
- What to Actually Do About It
- FAQ
The Real Numbers Behind AI Code Review
- Median time to first PR review is up 156.6%
- Average time spent in code review is up 199.6%
- Median time in review overall is up 441.5%
- AI code acceptance rate rose from 20% to 60%
Why Review Time Is Exploding, Not Shrinking
“We’re accumulating code faster than we are accumulating trust.” Kent Beck, “Trust Factory” newsletter, newsletter.kentbeck.com
The Benchmark Problem: Nobody Agrees What “Accurate” Means
| Tool | Vendor-reported catch rate | Independent benchmark result |
|---|---|---|
| Greptile | 82% (own 50-PR benchmark) | 24% (Martian benchmark) |
| GitHub Copilot | 54% (Greptile’s benchmark) | Not independently ranked in same test |
| CodeRabbit | 44 to 51% (varies by benchmark) | 46% (Macroscope’s ranking) |
| Cursor BugBot | Not separately vendor-reported | 42% (Macroscope’s ranking) |
| Macroscope | Self-reported top performer | 48% (its own ranking) |
What This Means If You Run an Engineering Org
“AI augments developer judgment; it can’t replace it.” GitHub Copilot code review product team, github.blog
“If an AI agent writes code, it’s on me to clean it up before my name shows up in git blame.” Jon Wiggins, ML Engineer, Respondology · via github.blog
The Contrarian Case: 19% Slower, Not 20% Faster
“I was complaining to people because I was like, ‘It’s helping me but I can’t figure out how to make it really help me a lot.’” Mike Judge, Principal Developer, Substantial · via MIT Technology Review
GitHub Code Quality’s July Launch: A Real Test Case
What to Actually Do About It
- Track rework rate and time-in-review, not just deployment frequency. A team that ships faster while rework climbs isn’t actually faster.
- Keep a mandatory human merge gate. This is GitHub’s own stated product philosophy, not just an internal best practice.
- Run AI self-review before human review, not instead of it. GitHub’s data shows this cuts trivial back-and-forth by roughly a third.
- Pilot any review tool against your own codebase’s known bugs before trusting a vendor’s published catch rate.
- Cap PR size. Multiple sources point to growing PR size, not tooling choice, as the actual driver of review slowdown.
FAQ
Does AI code review replace human code review?
How much time does AI code review actually save?
What is the most accurate AI code review tool?
Do AI code reviews catch more bugs than human reviewers?
Is AI-generated code more likely to have bugs than human-written code?
Where This Goes Next
Related Reading on NeuralWired
More posts
-
Anthropic’s Record IPO Raised $93.5 Billion, Then Its Stock Fell 8.5% in a Day as Trump Threatened a Data Center Tax
Anthropic’s IPO raised a record $93.5 billion and popped 34% at the open, then gave most of it back within a day. Here is what happened to ANT stock, why the market turned, and what Trump’s data center tax threat could mean next.
-
AI Hallucination in 2026: Why It Happens, What It Costs Enterprises, and the 5 Mitigation Strategies That Actually Work
AI hallucinations are costing enterprises $67 billion globally. This guide breaks down the root causes, the real financial and legal exposure, and the five mitigation strategies that measurably reduce risk in production systems.
-
US, China and Russia Issue a Joint Peace Declaration in Shenzhen: What It Says and What It Leaves Out
Trump, Xi and Putin sat down together for the first time in Shenzhen and issued a four-point peace declaration. It covers UN reform, energy, the ICC and nuclear risk, but never mentions Ukraine. Here is what the Trump Xi Putin meeting produced and what comes next.
-
Nvidia After Jensen Huang: Record Earnings, a Rejected Return and a Congressional Reckoning
Nvidia just reported $115.8 billion in quarterly revenue, and the board chair said Jensen Huang will not return as CEO. Here is what the Jensen Huang resignation means for the company, the China question and the congressional calendar.
-
Trump, Xi and Putin Will Meet in Shenzhen on November 19: What the First Three-Way Summit Means for Nuclear Arms
For the first time, the leaders of the US, China and Russia are scheduled to sit down together, in Shenzhen on November 19. The Trump Xi Putin summit arrives months after New START expired, and Beijing has so far rebuffed arms control talks. Here is what is confirmed, what is not, and what to watch.
-
Check Point Gateways Were Hacked Through PHP2Shell and a Root Exploit, Then a Wiper Hit the VMware Hosts Behind Them
The Check Point PHP2Shell attack began at a remote-access portal and ended with a wiper running on a VMware ESXi host. Check Point says only a couple of dozen systems were hit, but CISA has ordered federal agencies to check for compromise, not just patch. Here is what happened, what remains unknown, and which fix…
-
Zelensky Says the US Asked Whether Ukraine Would Leave Donbas in Exchange for NATO Membership
Official portrait of Ukrainian President Volodymyr Zelensky, published by the Office of the President of Ukraine under a CC BY 4.0 license. It is a 2019 file portrait and does not show the events described in the article.
-
Nvidia After Jensen Huang: Who Runs the AI Chip Giant Now, and What Congress Wants to Know
Jensen Huang’s resignation has left Nvidia with an interim CEO, an open search for a successor and congressional hearings on the calendar. Here is what is confirmed, what remains alleged, and what to watch when earnings land on November 17.
-
Will Europe Join the Next Ukraine Peace Talks? What Witkoff and Kushner Said After Meeting Putin
Steve Witkoff says Europeans will join the next round of Ukraine peace talks, but the Kremlin has not confirmed it. Here is who would speak for Europe, what Moscow wants first, and what could still stall the talks.
