Is the manual audit model not working?

AI audit tools have gotten seriously good lately, delivering solid results in hours instead of weeks. Meanwhile, traditional manual audits are still quoting 50-200K and 3-6 week timelines. Are teams actually still paying these rates?

The hybrid approach seems like it should be standard now: AI first pass to catch the obvious stuff, then targeted human review where it actually matters. We've been running tools like Sherlock and Guardix on our contracts – dozens of AI agents scanning DeFi-specific risks like access control, arithmetic overflows, and reentrancy. Gets you actionable insights in 1-1.5 hours.

Curious what others are seeing in practice.