Human review is a person actually reading what an agent produced and deciding whether it is good enough to keep. Not running it, not skimming it: reading it, understanding it, and taking responsibility for it. In an era where a model can generate more code than you can carefully read, this is the gate that still counts.
Why it does not go away
Everything upstream of human review filters for correctness. Tests prove behaviour, automated review flags the obvious, but none of it answers the questions that need judgment: Is this the right thing to build? Is the trade-off acceptable? Will future-me understand this? Those are human calls, and delegating them to a model that will happily agree with whatever you already think is how bad decisions ship.
It also anchors accountability. When code breaks in production, "the agent wrote it" is not an answer. Someone approved it, and human review is where that ownership is meant to live.
- Read the code, do not just check that it runs.
- Focus your attention where machines are weak: intent, design, the edge cases that matter.
- The reviewer owns the outcome, whoever or whatever wrote the code.
Related terms
Automated review
Automated review is putting a change through an AI reviewer before a person sees it, so a second agent flags likely bugs, missed edge cases, and smells. It catches the obvious cheaply, but it does not replace human judgment about whether the change is right.
Read definition →Human in the loop
Human in the loop means keeping a person in the agent's decision path to approve, steer, or verify its work. It is the deliberate counterweight to full autonomy.
Read definition →Vibe coding
Vibe coding is building software by prompting an agent and steering on feel, accepting the code it writes without reading every line. It is fast and freeing for prototypes and personal tools, and genuinely risky the moment the code has to run in production.
Read definition →