.NET Rocks! · Hands-On Engineering Podcasts · November 2026
Guest Anders Abel had said he is slow to use LLM coding for security components. Richard agrees and draws a line around code like cryptography.
…guy who said, look, just don't use any packages that are less than 48 hours old. Like the good guys will find weaknesses, but they don't find them instantly. So as long as you let those things marinate for a little while, your risk level goes way down. I don't know how you feel about that, Andrews.
…And also for security-sensitive things, it's always a good idea to be a bit reluctant, be a bit old-fashioned. And also, for example, LLM's coding, I've been very late or I'm still late on adopting it because I think that for security components, that's not maybe not the right place to experiment.
I'm not going to ask an LLM to generate a cryptographic library. That just seems like a bad idea. There are some deterministic things in this world that non-deterministic software should not touch. You
only live once, Richard.
I had this conversation with Philip DeRick, web security specialist. He recently got his first RFC published on best current practices for browser-based applications. And he said he used LLMs to generate security things. But the thing is that he knows the specs because he was part in writing those.…
Yeah, you need to understand the code it's generating when it's this sensitive, especially. I'm pretty loosey-goosey on the forms over data stuff. I'm not going to worry too much about that. But anything to do with authentication and stuff, like, holy man, be very. Very sure. And let's face it…