I used to review code the way most people do — open the PR, scan the diff, leave comments, hit approve or request changes.
Then I started something weird. I started taking my AI code review conversations and exporting them before I ever touched the actual PR.
Sounds unnecessary, right? Hear me out.
The Problem With Live AI Code Reviews
When I paste code into ChatGPT or Claude and ask for a review, the conversation is gold. The AI spots edge cases I missed. It suggests patterns I hadn't considered. It catches actual bugs.
But then the conversation disappears into the platform's history. And next time I'm reviewing similar code, I'm starting from zero again.
The Offline Review Workflow
Here's what I do now:
- Paste the code into Claude or ChatGPT and have a proper review conversation
- Ask follow-up questions — "what about concurrency?" "is this thread-safe?" "any security concerns?"
- When the conversation feels thorough, I export it as a PDF
- I read through the exported PDF like a design document before I touch the actual PR
- The PDF goes into my
reviews/folder for future reference
The export part takes about 20 seconds. I use XWX AI Chat Exporter because it preserves the code formatting perfectly — syntax highlighting and all — which matters when you're reviewing code.
Why This Works Better Than You'd Think
Reading an exported review offline is different from scrolling through a chat window. It feels more like reading a technical document. The table of contents (auto-generated for longer conversations) lets me jump to specific concerns.
And here's the thing I didn't expect: having a folder of past code review conversations means I can spot patterns across reviews. "We've had this same conversation about error handling three times now. Maybe we need a team standard."
That's a insight you can't get from ephemeral chat history.
The Compounding Effect
After two months of this, I have about 30 exported code review conversations. They've become a weird kind of institutional knowledge — even though I'm the only person on the team doing this.
When I see a new piece of code, I can quickly check: "Have I reviewed something similar before?" Often the answer is yes, and the exported conversation has notes I'd completely forgotten about.
Is This Overkill?
Maybe. If you're reviewing two PRs a week, probably. But if you're reviewing code daily and using AI as part of your process, exporting those conversations costs almost nothing and pays dividends the first time you revisit one.
The trick is consistency. Export every conversation that taught you something. File it. Forget about it until you need it.
Your future self — the one staring at a tricky PR at 4 PM on a Thursday — will thank you.
Top comments (0)