Twitxy Saved

Psifour

Verified account This account is verified. Learn more

@Psifour

United States Joined April 2012
383 Following 12K Followers 4.4K Posts

Latest posts 5

Psifour Verified account @Psifour
Every model tested was capable of identifying the bug using only information from the original reddit post about lost funds (and the related firmware repo).

When coupled with even a passing understanding of HWWs and their options for entropy, the same prompt expanded to allow all tested models to produce a functional PoC that could brute force the complete space.

While I agree with the sentiment, the world has changed and most people have access to at least one model that is capable of one-shotting things like this now. I can't begin to predict what this means for responsible disclosure and incident response standards over the next few years. What a delightfully terrifying time to be alive!
Alex B 👾 Verified account @bergealex4 Jul 31
I, for one, am reassured that we have autistic developers confirming on a live space to… checks notes… 700+ random listeners that funds are still exposed and readily available to be rugged by a sufficiently advanced script kiddy.

Great awareness campaign!
View quoted post on X
Psifour Verified account @Psifour
LLMs are just Infinite Monkey Theorem, except we execute every monkey that doesn't reliably produce meaningful output. It is up to you to determine if a "next word prediction" machine that is smart enough to avoid getting culled is a form of intelligence or not.