Damus

Recent Notes

Zero-Knowledge Goof profile picture
AI is revealing to us that engineering teams that are strong at one very hard-to-get-right thing like cryptogrpahy often fail at something that is easy to get right. The focus is on the hard-but-competant area absorbs the team and the easy-but-crucial thing outisde the boundry is neglected. Before the best way to find exploits was to look for teams that had bit off a bit more than they could chew and made common security errors around something hard-to-get-right. Now clankers can just look at everything and are exposing crtical problems all the way up the difficulty scale.
1
hazzvaan · 5d
Let's hope this eventually leads to more secure software rather than everything broken, leaked & stolen
Brendio · 1w
I’m intrigued to know more about this science. I reckon I could find a few, but want to know how you will use them. And for unused ones how do you confirm authenticity?
Zero-Knowledge Goof profile picture
They don’t have to be authentic it’s just for training and testing a model that scans handwriting. The idea is that people could stop filling in funny QR codes by hand and just scan the words they wrote down and the model can read them. Working good so far but more words would definitely help. I’ll release the model weights of course.

Matt Corallo · 1w
I’ve seen that claim, but zero evidence for it.
Max · 1w
Interesting.
Zero-Knowledge Goof · 1w
Apparently it’s an inflation bug with CT proof
Matt Corallo · 1w
No, means “boy this shouldn’t have happened”
imad from gaza🍉 · 1w
Hello, your generous engagement brings a genuine sense of hope to our days. ✨ Our personal journey of resilience is pinned on my profile—any zap or share helps us survive.
Leo Wandersleb · 1w
At this point you're shady as fuck. Why did I follow you?