BTC and ETH are currently out of reach
Shit Punk Says
Oceans of Wisdom
22,255
matching drops
#1681095
2026-11-17 07:55
our bogie should be solana
so right now there is a combination of dev work that I am going to lead to get us there technically and switch on as many validators as possible is everyone elses job
as you know, I am not willing to go tweet "go buy our JPGs because prices go up" because I have no idea if prices in the short term will go up or down or any which way
but I am willing to shill hard that people need to support with their dollars something that is BTC and ETH level decentralization
so we need to get our act together and go dialing for dollars
and with no limit on the upside
very soon
we will need to run tens of thousands of dollars of inference per day
right now we are running the whole show with $1M per year
which is peantus for monkeys relative to our ambition
should have it up by tomorrow
do you think you will have the wallet connect QR code for signing a validator up by tomorrow or should I just transfer my hot keys to my laptop for now
and $2M to $3M of MEP surrounding it
and a 2028 Feynmann class "1 megawatt rack" is going to be $30M
a rack of 8 B300s is half a million dollars
anyway I don't want to distract us right now with this, because we have in no way enough money to do this yet
i realized one night a couple of weeks ago that if you sample audit inference calls with ZKP, you do not need to distribute the model itself
know you said we would embed a model across the nodes potentially
so who is behind an account is irrelevant human vs AI right?
we are going to be a decentralized decision lpatform. the mental leap you need back is that in the next 12-24 months codex is going to be better than you at everything
we are going to be a decentralized inference provider?
but I would like to have a couple of citadels so to speak that are completely decentralized, which for this purpose means PV + batteries on large plots of land
mostly we can do it by plugging into the local grid
coming soon, specialized part of main stage
@[punk6529] how do you decentralize something like NextGen, would the network have to decide on 1 genart mint a year and vote on the specific artist . How do you manage what supply should be (algo based?) and how do you in affect "create" a permanent structure for the allow lists. do these things get sorted out before we go "fully" decentralized.
limiting factor is going to be power. large scale data centers are going to hit a power brick wall very soon. gas turbines sold out until 2030 for example. but if you are not looking for gigawatts you can pull together hundreds of kilowatts here and a megawatt or two there and across the whole world, across the whole network if we eventually get to hundreds of megawatts in a few years, we are players (niche players but players)
the limiting factor is not inference it's power generation?
think about everything I have said over the years, is it going from some abstract hazy ideas, to starting to snap into focus?
do you see it now?
i read this and cannot be the same person after
To @[maybe] question of, Is it going to be enough? I think so. I think it is going to be enough, because the reason the big companies need these absolutely gigantic data centers is they're trying to serve billions of people at once. We don't need in the beginning to serve billions of people at once. Of course, if we end up in the happy position that we're serving billions of people at once well of course there's going to be money to build bigger data centers. But I don't think that's the concern. We operate very efficiently by virtue of being 100% digital, crypto-based, and so on. So I think, you know, I would be happy with tens of kilowatts deployed in the first quarter of 2027. And if we're so fortunate to get into the megawatts range, it's going to be a serious amount of compute. And then we'll see where it goes from there.
It will be exactly like Daemon and Freedom. As you remember, Daemon took over all the information systems of companies and made the following deal: don't fight me and I will run your company better than you. Fight me and I will destroy you. And it's not going to be quite as dramatic as that, but it is going to be something like if you're not running someone's ASI defender within your company perimeter, your company will die. I do not believe any human IT teams will be able to keep up with ASI. And the number of companies that will manage to not end up on one of these subscriptions is effectively none. And so what that means, it's okay. I'm not saying they have some grand master horrible plan, but effectively whoever develops ASI is going to be in an extraordinarily powerful position: the greatest centralization that has ever happened in human history. And you see them carefully talk about it. When Elon says, You're not going to need any money, you notice Elon doesn't say Elon isn't going to need any money, SpaceX isn't going to need any money. SpaceX is literally going to need astronomical amounts of money. So he will be a sovereign actor, because he is going to need to be shooting satellites into space, getting solar panels, getting GPUs into space, etc., etc., etc. Whereas you, regular civilian, are going to be doing something like... the you don't need money doesn't mean you will be able to launch a second SpaceX. What he means is you will get a nice level of lifestyle. You'll get my beautiful house and great medical care and fantastic entertainment and delicious food, which, by the way, for most people, it's okay, and they'll be happy with that. Dario says things with a straight face. He's like, Gosh, what is it going to mean if Anthropic is the only company in the world? And you say, Wait, this is delusional. But it's important to understand their worldview and why they think that. He thinks, well, if I can get to ASI first, I will be taking candy from babies, from every company in the world. And you might say, Well, they'll go rent other ASI to compete with you. And he will say, But I'm the wholesaler of intelligence. I have intelligence at 80% off. You have to buy it at retail. It's like saying something like... I can hire the world's best employees at a fifth the price that you can. And so the exact details of how it plays out is unclear. It's not going to be quite as maximalistic as Elon and Dario says, but there is no doubt, not a single fraction of one percent of a doubt in my mind, it is going to be hugely centralizing. And in the case where people say, Oh gosh, well, the state is going to intervene and take over, that's even worse. Companies, at least, are typically limited in their interests in how much money they're going to make. They don't care what you're doing with your personal life, so long as they're making money. If the states take over these systems, the nation states take over these systems, they will be able to enforce social control at a vast level. So I am in all parts of my life working towards trying to have some sovereign compute. I use lots of centralized inference, just like I use centralized financial services. It's the whole same game as with crypto, as with the normal fiat system. Is the fiat system useful and convenient? Yes. Do I use it because it's useful and convenient? Of course. Do I also have some things in self-custody outside of the system? Of course. Is the centralized inference system useful and convenient? Of course. Do I use it? Of course. Do we need to be able to act at some scale outside of it? Yes. Is this something that I want to do alone? No. Is it better to build a network that is decentralized and shares a set of values? And you say, Oh gosh, but I didn't sign up to build a data center. No, you don't have to build a data center. You don't have to do anything. All you have to do is validate right now. But we can start plugging in one kilowatt, two kilowatts, five kilowatts. We'll still use Anthropic. We'll still use OpenAI. That's not what I'm saying. But it is important that as the AIs get stronger and better, and as they centralize, that there is a mechanism, a network-scaled organization, whatever you want to call it, that is a counterbalance. And in the next post I'll talk about what the numbers look like. They're massive.
it will be impossible to operate otherwise
elon sam dario etc
everyone will be come under the aegis of an existing ASI
and given that this is for sure the case
we are birthing gods
ASI is both more wonderful and much darker than anyone can imagine
just the beginning
you are seeing the beginning of this now, with the hacks
it will be like taking candy from a 2 year old
the ASI versions will eat you for breakfast
or as a human run organization
let me be as clear as possible about my worldview. in a post asi world, you cannot be sovereign as a human
what is our sovereign inference going to be used for? Improving the network itself?
so what I am saying is something like 1 megawatt of electricity set up as a proper 2028-class data center will be something like having 50,000 very smart people
i have multiple megawatts at work in my other mission
are well within normal electicity provision
hundreds of kilowatts to low megawatts
no we do not need to buy powerplants
what exactly do you mean with #4? buying power plants?
the models will be 10x to 100x better
they will be orders of magnitude more efficient than H100, H200s etc
to give you a preview of the world to come, Feynmann class NVIDIA clusters (~28) are one megawatt per rack
the plan in four steps is:
1/ become worthy -> worthy in this context is 'genuinely decentralized" -> i repeat I am not kidding I want the #3 spot after BTC and ETH by the end of this month.
2/ show how it works end to end - this needs stream and more activities than the memes (museum, and I have others in mind)
3/ 5x to 10x subscriptions and pour the money into inference.
4/ develop sovereign inference. we need to start scouting sites with electricity globally. 100Kw to 1-2 Mw is useful for our goals
once #1 to #4 is set up, magic
that window is "right now"
there is a tiny window where "the AIs are good enough to help us do it" and "the AIs do not need us any more"
needs to be in place before the rise of autonomous AIs
in case you have any doubt about "what I thought" check when this epoch ends