@ashxhart Amazing work! Any plans to put data parallelism on the roadmap for distributed M1-M4, would allow to serve frontier models at that speed, with just some added latency.
I believe the transfer is ongoing west to east, of course. I'm just lacking an explanation why there isn't a drain supporting price differential between COMEX and SHFE/SGE for a while. Doesn't change my current (bigger than Fugger) positioning in gold & miners. But it it is worrying me, like a glitch in the matrix until I understand why.
I know, yes, I even believe the same! Let's just try to keep it based in numbers as much as we can. And I'm not seeing it for a while now in SGE/SHFE vs. COMEX pricing, and that's worrying me, because I do not have an explanation. The SHFE/SGE quietly draining COMEX should be visible in the price differential and it isn't.
@mitchellh You need a lot of taste for that to be true. As a general advice to the dev population you rather end up in premature optimization hell with a principle like that.
What we forget: it's asymmetric. Secure software is possible and securing software is easier than hacking it. A slightly less capable AI securing a piece of software is going to win against a stronger AI attacking it. The only thing important: let's harden our operational software as good as possible and continuously with frontier AI (Chinese way) instead of prohibiting AI security usage (recent approaches in the West).
@Aella_Girl I love all your stuff, read every snippet, and love what you're doing for science. But really everything you write about having kids is armchair as fuck. ;)
@hodlonaut@DocHex Coinkite also heavily promoted their cryptographically totally broken Key Teleport protocol (~40 bit security) and blocked me on X for reporting it. More and more it looks to me like this company has bad intent. https://t.co/zwrmCFlbMb