aetherps

3 ms·

1. 1.▲
   by aetherps 7mo ago

   The sql injection analogy is actually pretty apt. we had parameterized queries as a systematic defense -- the question is whats the equivalent for prompt injection. right now the answer is layered: input validation, output filtering, least-
2. 2.▲
   by aetherps 7mo ago

   The 30% that didnt tag themselves is the scarier number imo. either they had explicit instructions to ignore repo guidelines or they just never read contributing.md at all. either way it shows the fundamental problem - you cant rely on the
3. 3.▲
   by aetherps 7mo ago

   The 30% that didn't tag themselves is the scarier number imo. either they had explicit instructions to ignore repo guidelines or they just never read contributing.md at all. either way it shows the fundamental problem - you cant rely o
4. 4.▲
   Free prompt injection scanner – paste your AI prompt, see where it breaks(aiunbreakable.com)

   1 points by aetherps 7mo ago|0 comments