The volume of machine-generated code flooding development teams has created a crisis in traditional code review workflows. When asked how to manage this surge, experienced engineers offer sharply divergent solutions. The conversation about what comes next remains far from settled, and two prominent voices in the field will tackle the question head-on during a live event scheduled for Tuesday, September 29.

John Bristowe, Principal Developer Advocate at Octopus Deploy, and Viktor Farcic, the platform engineering expert behind DevOps Toolkit, will participate in a public discussion titled "Human Review vs. Verified Pipelines: What Catches Bugs in the Age of AI Code."

The Scale of AI Adoption and Its Consequences

The numbers tell a stark story about how quickly AI has become embedded in development workflows. The 2026 DORA report shows that 90% of developers now rely on AI tools at work. This widespread adoption has fundamentally altered the pace of code production: developers are now merging 98% more pull requests than they did before AI became commonplace.

Yet this acceleration has come with significant quality trade-offs. Defect rates have climbed sharply, with bugs per developer rising 54%. An examination of 10,000 developers revealed an even more alarming trend: incidents per pull request have jumped 243%. According to Octopus Deploy's AI Pulse report, the problem stems from how AI coding agents operate—they generate large code blocks that human reviewers cannot realistically comprehend in their entirety, even though faster code creation remains one of AI's primary benefits.

The Bottleneck Shifts, Not Disappears

Development teams have embraced automated code generation far more readily than they have adopted automated code review. This mismatch has simply relocated the human bottleneck rather than eliminating it. Additionally, AI-powered code review tools may suffer from the same limitations as the agents that generated the code in the first place.

Bristowe contends that code review has essentially become performative. A human cannot meaningfully audit a 40,000-line pull request created by an AI agent, particularly when that reviewer had no involvement in the reasoning process behind the code and cannot realistically understand every change it introduces.

An Alternative: Policy-Driven Pipelines

Rather than asking humans to review more code or deploying additional AI systems, Bristowe proposes shifting the quality checkpoint from human reviewers to the deployment pipeline itself. This approach does not necessarily require more AI. Instead, he advocates for embedding strong "policy-as-code" rules into deployment standards that flag only violations of those policies. Humans then focus on handling exceptions rather than attempting to review entire codebases.

Farcic is expected to challenge this framework, questioning whether policy-as-code can truly capture nuanced judgment and whether the approach simply creates another layer of accountability without solving the underlying problem. The discussion will also address the impact on junior engineers, who can no longer expect to learn from teams of humans writing and reviewing code together.

Event Details

The live conversation begins at 2:30 p.m. Eastern/11:30 a.m. Pacific on Tuesday, September 29. Attendance is free, and those who participate live will receive a companion resource based on data from Octopus Deploy's AI Pulse report. Registration is open now.

Key Topics

  • How AI-generated code has undermined the foundational assumptions of traditional code review, and why simply conducting more reviews cannot solve the problem
  • The limitations of using AI to evaluate AI-generated code, including shared training data and identical blind spots
  • Building deployment pipelines that validate every release against a predetermined ruleset, regardless of whether humans or machines wrote the code
  • Identifying where human code review remains valuable and where pipeline automation should take precedence

Source: The New Stack