Engineer argues agent-written code still needs human ownership
Original title: Here Be Monsters: What Do You Still Need to Know When Agents Do the Work?
A SonarQube complexity flag on Codex-generated React code led the author to inspect, then have the agent split one file into ten.
The author describes CI failing because SonarQube flagged cyclomatic complexity in App.tsx. Codex had written about 15 components on single lines; after formatting, the logic was correct but dense, so the agent refactored it into 10 files and CI passed. The essay argues that as agents write unseen code, engineers must still know enough to spot when something is wrong and dive in.
Unconfirmed: This is one engineer's anecdote and opinion, not a measured study of agent failure rates.
- App.tsx was about 30 lines, with roughly 15 components on single lines.
- The author found the complexity matched requirements and the logic was correct.
- Codex split the work into 10 files, after which CI passed.
- The essay contrasts optimist and skeptic views of autonomous coding agents.
From the original article
12 min read at blog.adrq.dev
CI was failing. This was unexpected. I have dialed in my agentic workflow to a point where I would be surprised if one of my agents committed a branch that had failing tests.
By adrq · blog.adrq.dev
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.