A 90-Minute Workshop for the New Reviewer Role
Monday, 9:14 a.m. A pull request lands with the label generated . Four hundred lines, a new endpoint, and a test suite that passes. The reviewer opens it, reads the first function, and feels the familiar squeeze: the code is plausible, the tests are green, and nothing looks wrong. That is the moment the job changed. For years, the hard part of software work was writing code. The bottleneck has…
A new role in software development is the reviewer of AI-generated code, but few receive training for this task. To help new reviewers, a 90-minute workshop has been designed. The workshop uses a simple lab that can run on free infrastructure, making it accessible to teams with limited resources.
The lab requires a place to run tests and a way to generate sample pull requests. MonkeyCode offers free model access and a free server option, keeping the cost at zero for small groups. The workshop focuses on teaching a checklist for reviewing generated code, rather than specific tooling. This checklist helps reviewers identify common defects in generated code.
During the workshop, students go through four exercises. The first exercise involves reviewing a pull request using a checklist. They mark a five-item checklist, and most people find no defects on the first pass. This activity teaches that unaided review of generated code is mostly pattern matching and often misses edge cases.
The second exercise involves writing a failing test. Students write a test that should pass if the code is correct, but it fails. This exercise shifts the review from opinions to evidence. The third exercise involves cross-grading the reviews. Pairs swap checklists and grade each other against the planted defect list. This process makes review quality measurable.
The workshop ends with a debrief, where the facilitator reveals the defects and asks which checklist items caught them. Usually, one item catches all the defects. The final part of the workshop is a wrap-up, where students take the checklist back to their real reviews.
The first sample pull request in the workshop is a function that merges overlapping intervals. The code looks clean, but the checklist makes it less comfortable. The checklist highlights several important items for review, such as empty input, mutation of caller data, contained intervals, return type, and error behavior on malformed input.
The second item is the trap, as the function mutates the caller's list. The real bug is that the code overwrites the end instead of taking the maximum, which a generated test suite often misses.
The takeaway is that the reviewer role is new but trainable. Ninety minutes of focused training on planted defects can help new reviewers see generated code as puzzles rather than magic tricks. The workshop repo runs on a free server with free model access, making it accessible to teams with limited resources.
Written by urgent.news from Dev.to's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
