I build AI agents professionally. I used Passmark on my personal tracker at 4 AM. It found 8 bugs — 5 invisible to any manual test.
It's 4 AM. I have a daily tracker app running on localhost, a fresh Supabase test database seeded with fake data, and a Playwright suite I spent the last few hours writing.
I'm not nervous about the results. I use this app every day. I built the gym module with a paranoid little EXIF guard that rejects photos older than 6 hours. I wrote the XP logic myself. I know where the rough edges are.
The first test fails in 37 seconds.