How to review AI code professionally?

AI generates just a ton of code these days. Way more than I can actually read.

On a normal day, I have 3 to 5 agents running in parallel (I wrote about how I work with them in my post about Agentic Mode). And every one of them ends with a diff I have to review.

Writing code got cheap. Reviewing it didn't.

So… I needed a better way to review all this code.

For the last three months, I've been using one tool for this every day. It's called tuicr.

tuicr is an open-source code review tool for the terminal, written in Rust. It works with your local changes and GitHub pull requests.

You leave comments right on the diff, and your AI agent gets them as a numbered list, each one pointing to a file and a line. No more copying code into the chat by hand.

It's not perfect, though. I'll show you its drawbacks and a few alternatives at the end.

The idea is simple: the review lives in the same place as the agents. In the terminal.

I go through the diff, leave a few comments, and press y (or run :export). The whole review lands in the clipboard, and I paste it back into the agent. That's the whole loop.

And when I review a GitHub pull request, :submit posts the same comments as a real PR review, right on the lines.

1. Vim-like navigation: jump between changed files, hunks, and your comments! And if you prefer, you can use your mouse too.
2. You can review multiple revisions at the same time!
3. You can open the current file in your editor right from tuicr.
4. Rich commenting: you can comment on a whole file, a range of lines, or a single line.
5. You can mark files and hunks as reviewed, and tuicr remembers it between sessions. If an agent changes a file again, tuicr unmarks it, so I never miss new changes.
6. Comment types: I mark a comment as issue, and the export tells the agent it "must fix before merge". The types are defined in the config, so you can add your own.

1. Github pull requests comments don't support synchronization (only publishing in one way)
2. There is no world level diff highlighting like delta has
3. Some times there is no synchronization with actual revision, or you can't open tuicr for autodetect changes in the current revision
4. It doesn't switch to dark mode on the fly. It follows the system theme only at startup, so I have to restart it (Yep! It's still important to me 🙂)

Tip: if tuicr ignores your system theme, remove theme from the config and set theme_dark / theme_light instead.

If you live in the terminal, I think the answer is yes.

If you spend most of your day in Neovim, it's a bit different. review.nvim gives you the same loop right in the editor: comments on a diff, exported as markdown for your agent. It was even inspired by tuicr.

For GitHub pull requests, there is octo.nvim. But that's two plugins for two jobs, and tuicr does both in one place.

In Emacs, I use reviewer.el for local review. It annotates any buffer, including a diff, and renders all the notes as an org document I can hand to an agent.

But for GitHub pull requests, I haven't found a convenient solution in Emacs yet.

There are some alternatives for review AI agents tools

https://github.com/agavra/tuicr#how-it-compares

But, to be honest, almost everything works better in tuicr