• 0 Posts
  • 13 Comments
Joined 5 months ago
cake
Cake day: May 25th, 2026

help-circle
  • Depends on what they are lying about. If they are lying about using it responsibly then they are goons, but truly responsible use can be safe. Oversimplified example: a for loop printing all digits in order is just as good if generated by most models as when written by hand. How much you can realistically extrapolate from this example depends the model you use and the scrutiny you apply to the output. Now the ethics and implications of using LLMs are a different matter but it’s not a black and white issue for sure. IMHO after over a decade in the industry.






  • You’re probably right that the reality of proton gray and not so black and white and they have done both good and bad but on pentesting vibe coded code with just more LLMs is wrong and probably why you got downvoted so much. You absolutely have to have a human expert involved in security, there is no alternative. Would you and your dearest family fly on a plane designed entirely by AI and checked exclusively by AI? I didn’t think so. LLMs are still a great tool and an amazing second set of eyes but they can’t be trusted




  • Oh god yes, I can’t tell you how many times i’ve seen execs in general come in and confidently peddle their motivational crap and ethos only to be proven wrong 2 weeks later and having to backtrack and reframe their messaging with absolutely no consequences. But they only do that from peer pressure not because they actually believe they’re wrong, often exactly as the proton ceo did, saying they still want to push forward just in a different way. These people are unembarrassable and you can only get away from them if you quit or wait for them to quit and be fired, there’s no reforming.