AI products tend to fail confidently on the inputs their builder never imagined, so I'd rather find those inputs myself before real people do. This thread runs from handing Launchling French pirate founders in 2025 to a fresh session trying to disprove the top findings of Fold's first audit with runnable reproductions.
Breaking it on purpose
How do you find the ways an AI product fails before your users do?
adversarial testing · red-teaming · edge cases · prompt injection
Start here: Satirical Startups & French Pirate Founders: How Launchling Handles Absurdity by Design
Builds
- EXP-006Building · Pilot
Fold's review panel
Never mark your own homework: independent AI reviewers for a solo builder
Independent AI reviewers for an app I was building faster than I could read its code. They caught bugs that had passed the tests, and one finding changed what Fold is.
Notes
- Satirical Startups & French Pirate Founders: How Launchling Handles Absurdity by Designadversarial testing · red-teaming · edge cases2025 post