@steadmanticore@infosec.exchange
Programmer, ‘Elite Hacker’ (HTB), and former captain of #YegSecCTF. I will not apologize for my nerdy dad jokes. Team #ds9. Tweets are my own. He/him.
I've thought about it more and I think another problem with LLM non-determinism is the "prompting" problem. You can rephrase the prompt very slightly and bias the model in a completely different way. The randomness in the output for the same prompt is one thing, but the randomness in the input is a completely separate problem and can't be ignored. It leads to a lot of mysticism about "oh yes, I am a prompt master, only I can prompt properly" (true or not).
Thanks for your interest in trying out Elk, our work-in-progress Mastodon web client!
Expect some bugs and missing features here and there. we are working hard on the development and improving it over time.
Elk is Open Source. If you'd like to help with testing, giving feedback, or contributing, reach out to us on GitHub and get involved.
To boost development, you can sponsor the Team through GitHub Sponsors. We hope you enjoy Elk!
