I Had Claude and GPT Review Code, and the Difference Is Pretty Big
I've tried both, and Claude's refactoring suggestions are more useful, while GPT is a bit less prone to making up nonexistent libraries.
I'm curious what you use as your main tool at work.
5 answers
Facts, Claude really nails refactoring suggestions. But it still sometimes makes up methods that don't exist and causes compile errors, same as ever lol.
Got a source? If you're just going off vibes, that doesn't really mean much.
In practice, I keep both open and use them side by side. For reviews that involve reworking the design or architecture, I run them through Claude; for questions about new library APIs or documentation-based questions, I throw them at GPT. In the end, rather than standardizing on one model, splitting them by the nature of the task puts less stress on me. The costs differ for each, and the context lengths are different too...