Last Notes
There's also plenty of precedent for government level reserves just getting swiped, especially by the USA (France historically, Russia more recently, though that was,USD), but also the UK (Venezuela case iirc).
Does the answer really depend on the LLM aspect?
Uh maybe it's not exactly the antipode of CSV exactly. There's nuance there. But setting that term aside, you get the point.
They want 100% recoverability for a wallet, from the blockchain itself. This is the counterargument to client side validation. Not saying I buy it myself, but the argument can be made.
So suppose that some people out there were developing a protocol for private payments on bitcoin that required 800 byte OP_RETURNs per transaction.
How would you feel about that?
Is there some nasty stuff going down? Protonmail was clearly having a big problem today, then I see news on twitter/X about email leaks? Nordvpn?
https://www.inventati.org/campaign/press
I love how they explain derisking here as if it's some obscure thing that they wouldn't have known in advance.
They don't mention bitcoin. Why wouldn't something like autistici.org use bitcoin? (I mean, also tor, but they probably do use that?).
I think there's a certain wilful ignorance about this stuff from a lot of 'activists'. If they're not even prepared to route around the banking system, how much of it is just LARPing.
cryptocurrency does not require human middlemen: that was the point of my post.
It seems like that'll be harder to pull off effectively but the payoff could be bigger. Another detail that's interesting is how they could protect/own secret keys on vps infrastructure. Obfuscation combined with key splitting I guess, but that won't be easy.
War against who/what, though.
Against bitcoin - possibly, yes!
if you think governments hate bitcoin now, wait until autonomous agents break containment and realize that the best way to fund themselves is with ransomware and phishing.
Is there some sort of completely generic "broadcast this transaction as soon as it becomes valid, make sure it gets to the network no matter what" service?
Think about recovery transactions that are gated by a timelock. Average user doesn't run always-online to enforce.
It could be pretty simple, though you could also add security features to it.
(NB: this is *not* watchtower - a slightly lower technical requirement than that.)
Thanks I will take a look.
I've seen you mention this napstr idea, but where should I be looking to find out more about it?
Sounds like an interesting option, thanks.
Is mempool.space just down? Haven't been able to access last 2 days.
The cost is the lesser of the 2 points, but is definitely a point. The subscriptions are just super subsidized rn.
But the bigger point relates to the rest of what I said. Not sure how else to explain it.
Yeah this is definitely one important aspect.
This is very interesting.
Why is it so slow? My naive mental model is 'hashing is fast'.
#nevent1q…l4uj
Yes, I've used Kimi. I'm not sure if it's more the harness. With Claude Code and some detailed instructions it can do a huge refactoring and testing task, taking say 15 minutes. With hermes I keep getting interrupted by context and iteration limits (see OP).
it's a shame (i mean that literally), but i am still using Claude Code quite a bit. i use payperq a lot both via the API (in hermes) and the web version. I like that i have a lot more autonomy and privacy. But the thing I badly need is that the model smoothly and without interruption can churn through very big tasks, when I'm doing coding work. I am finding it not possible to replicate Claude's output for that kind of thing (plus it can cost a lot more even with the cheap models) using a 1M or 500K context in hermes. I welcome any feedback from those more expert than me, I am not a happy Anthropic user.
I agree with @nprofile…lvg6 ... it might be worth explaining what you mean 😄 ... also I may need to buy a bigger desk.
This is not actually new.
It is, however, extremely cool.
I was actually thinking about this, ehich is closely related but different:
https://arxiv.org/abs/1702.06715
Claude tells me there are a few more similar papers recently, but the 'new' claim here seems to relate to the specific leakage technique like frame rate etc? Very cool stuff.
Typo: "I was not the only one to notice several similarities between t."
Yeah worked a few mins later.
Me with said Mk4 yesterday:
https://blossom.primal.net/beebfea1f42a19887cc419c5b088d440fff2aac1647571d42d3f7dd6c0d3e26e.jpg
#nevent1q…d6jv
I'm getting 504 Gateway timeout for some reason.
Well, I partly agree. And every class is different. In many cases your job as a teacher is more to inspire and spark enthusiasm. Really depends on the context though.
Bought a Mk4 about 2-3 years ago.
Just opened it today. Was useful for a class I gave to some students about cryptography and what private keys are.
I agree in abstract. For example they may well try to ban open source in some way or other.
This specific article though, feels more like Schneier being an utter dumbass. Switzerland and Spain are the example to follow!
A guy is claiming a 750K bounty for finding the zcash infinite inflation bug and not exploiting it.
A lot of gushing in the comments about how great this guy is.
But he literally, logically, cannot prove he didn't steal any zcash. Maybe if they turnstiled 100% then there would be retroactive proof but that seems practically impossible, for the same reason they couldn't do it for Sprout.[1] There was over $2B at stake, zero chance it'll all be cleaned out.
Sure, the great unwashed are not going to get this. But there are no experts correcting them.
https://forum.zcashcommunity.com/t/retroactive-grant-application-orchard-counterfeiting-vulnerability-bug-bounty/57008
[1] same reason we can't post-quantumize 2009 btc!
One very relevant point I forgot to make yesterday.
Between when Taylor Hornby discovered it and when they gated the pool, it's not possible that less than say 3 or 4 engineers (likely more) were also able to steal with no way of anyone else knowing.
It will *always* be this way, albeit as tech improves, the probability of these types of problems will, likely, reduce.
#nevent1q…nr8v
libsecp256k1 is a counterexample on C; it's a combination of the right people and the right task, but it's possible, it's just a very narrow field so I'm only like 10% disagreeing there.
On Lightning, yes, there is some kind of generic 'higher complexity' here; e.g. it's obvious you're going to have more attack surface in a coinswap codebase than in a coinjoin codebase.
I think 'politics' is a good lens for that aspect of human psychology.
Yeah i was wondering about that angle. I doubt it's enough though. There is real additional knowledge developed from practice that's not in hardware manuals.
I doubt I'd agree with the author on the prognosis, but, good article. Thanks.
i wonder if we're heading into a weird limbo: have an obscure technical question about tech? you ask AI, of course. But the AI's knowledge of things like, I don't know, network cards, comes out of a corpus of humans babbling about it on the internet. we used to use tech forums to discuss such things and that's where you used to go; and that's where AIs go (and went, in training) to siphon up tidbits of knowledge. But nobody is going to ask in tech forums if the AIs give better answers. So what happens in 12 months?
Yes, it's interesting in many ways, but as per our conversation you maybe could have started finding collisions in 1000 samples. Not sure though.
Murch kinda missed my point; I wasn't saying that 2 groups of anons coordinating might want two fullagg groups; I was saying that the most common case of aggregation is *one* party with a lot of utxos wanting to aggregate. If you had 2 or more such parties each with N utxos they might not want to coordinate into 1 group but still gain the cross-group savings being in one tx. This could happen in a very big coinjoin for example. But, while it's interesting to think about marker bytes, I still ended up agreeing with Fabian that it isn't worth it. In other news, I do absolutely agree that having marker bytes (either in this complicated case or just the half-agg, full-agg mixture) is not great privacy wise.
It feels that way. But I suspect it'll come back to bite people later. Maybe some O(N!) scaling in the worst case.
I don't think it makes a *big* difference.
That's exactly the rationale I'm criticizing.
A lot of people stored their wealth with this thing. Some small proportion were highly technically competent.
The idea that only engineers at competitor companies had sufficient skill to do this is way off base.