I've been evaluating tools for my team's procurement pipeline. After running a few through real PRs, I'm not impressed.
Most comments I see are just restating basic style guides or pointing out simple syntax. Things like "Consider using a more descriptive variable name" on a loop iterator. Or flagging a missing null check that any decent linter already catches. This isn't "AI review," it's a noisy wrapper around existing rules.
Where's the actual analysis? I need to see comments on logical flaws, potential race conditions, or subtle security vulnerabilities that a human might miss. So far, the signal-to-noise ratio is terrible.
What tools are you all using that actually provide high-value feedback? I'm looking at concrete metrics: reduction in human review time, not just volume of bot comments.