Hacker News Reader

Might happen after LinkedIn's built-in AI integration goes live.

The cat and mouse game begins, if a LinkedIn engineer sees this post there's a decent chance they try to filter out these sorts of prompt injections

I don't think prompt injections are a solved problem yet.

And they won't be in the near future. We don't have a good understanding of how to stop them.

Maybe compare the output to the expected format and see if it looks like it has been manipulated?