Show HN: adamsreview – better multi-agent PR reviews for Claude Code

A new Hacker News demo called adamsreview promises to turn code checking into a full-blown AI assembly line: multiple AI helpers inspect a code change from different angles, another pass double-checks the findings, and an automated fix loop can even try repairs and undo bad ones before saving anything. The creator says it catches more real problems than several built-in and paid alternatives — though, to be fair, the evidence is the classic internet disclaimer: “anecdotal, n=me.”

But the real fireworks were in the comments, where readers reacted less like impressed shoppers and more like people watching someone build a Rube Goldberg machine to butter toast. One commenter squinted at the process and called it a “fair bit of ceremony,” while another dropped the line of the thread: “Holy vibe coding batman” — basically accusing the project of being a giant pile of prompts stacked on top of prompts. The biggest existential crisis came from a user who joked, “I pay Claude, to use Claude, to write instructions for Claude, to review code from Claude,” which instantly became the unofficial theme of the discussion.

That’s where the split emerged: is this a clever way to make AI safer and more useful, or a sign that software people are now fighting machine-made complexity with even more machine-made complexity? Another commenter said AI agents reviewing AI-written code felt “dirty,” especially for serious production work, and pitched a human-in-the-loop alternative instead. So yes, the tool is about better code reviews — but the comment section made it about something juicier: have developers started outsourcing so much judgment to AI that the whole thing is becoming absurd?