VascularHand

GFYS

[VascularHand @VascularHand]

@_SrGio2 @Pirat_Nation yep :(. Next thing you know that old 386 in your mom's basement will be worth $2k. can't win.

[VascularHand @VascularHand]

Seriously, just give us a way to call pro directly from codex, even if I have to purchase credits to do so. It's a huge pain to copy and paste between the web and codex client. And on the web pro doesn't see my private git repos reliably. Please, do this. I can get fable in Claude Code. Why Can't I get pro in codex?

[VascularHand @VascularHand]

most as in who exactly? Most of the earth's population, because that feels overly optimistic, by far. He can say whatever he wants as often as he wants. But if he wants people to actually listen, which apparently they don't or he wouldn't speak about this as often as he appears to, he should create something that displaces LLMs. That's the message people will listen to.

[VascularHand @VascularHand]

@disclosetv oh thank god there's nothing a summit can't fix.

[VascularHand @VascularHand]

@moskstraum21745 @digitalix @Prince_Canuma @pupposandro I honestly don't know, I've never tried to do anything like that with it. It does have a full IPMI system, so if you wanted to turn it on or off remotely it would be trivial to do.

[VascularHand @VascularHand]

bro, you need to step back. Why would you think: "access to "frontier" models was cut off, and presumably remains cut off forever" ? Either you have AI Psychosis or you aren't really thinking through things deeply. What happened was not a signal that the model "presumably remains cut off forever". https://t.co/bnYh5jZupP

[Chubby♨️ @kimmonismus]

It is absolutely crazy how the last two weeks have changed the entire future.

It is unprecedented that access to "frontier" models was cut off,and presumably remains cut off forever.

It feels like a watershed moment, as if access to the highest level of human intelligence had been blocked.

Open source is the future. Open source is the solution. The last two weeks have powerfully demonstrated this.

[VascularHand @VascularHand]

@__tinygrad__ This is such a dumb take

[VascularHand @VascularHand]

@moskstraum21745 @digitalix @Prince_Canuma @pupposandro I have one

[VascularHand @VascularHand]

@moskstraum21745 @digitalix @Prince_Canuma @pupposandro This board can probably do what you want and the CPU pricing isn't horrible: ASRock SIENAD8-2L2T

[VascularHand @VascularHand]

@kimmonismus Please give it time. You are "catastrophizing". Between now and the end of our lifetime a lot of things can happen.

[VascularHand @VascularHand]

@Vladimi39901452 @kimmonismus This. In terms of making our own, if that becomes reasonable then look for the hardware itself to be restricted, or the breakthroughs that allow less powerful systems to create models like Fable will end up being classified.

[VascularHand @VascularHand]

You are not thinking about second and third order things @kimmonismus. When the models become capable of being used as powerful weapons, no government on earth will allow them to be given away. Just because they released GLM 5.2 doesn't mean they would release GLM 7. Why would the Chinese government release something like that? There comes a time when the benefit of releasing a model is eclipsed by keeping the capability of that model for yourself.

[VascularHand @VascularHand]

you had to know that this was going to happen eventually right? The models could get dangerous. It feels bad to be in this moment, but it was going to happen, and what is most important is what happens next, how the government handles what can be released and to who.

Everyone thinking that open source models from China are going to save us are not thinking clearly. When the model itself becomes a strategic advantage, do people really think China is going to give that model away?

The government also sees these models as potential weapons - they aren't in the business of giving weapons away, they keep those weapons close to their chest and secret. Only the China Labs and the Chinese Government even know if there are already more powerful models than GLM 5.2 at this point.

This is not the real issue either. The real issue is what happens if we find ways to create better models with less compute, and the hardware itself that could be used to make a powerful model, say like Fable, with $100,000 worth of compute and a bunch of AI agents running locally? Now many companies, probably millions of them, could do this.

The real question that is haunting the labs and the government is how long until more people have the capability to build these models.

[VascularHand @VascularHand]

@AndrewCurran_ what makes you think China will allow extremely powerful Chinese models to be released at all? They will have an incentive to keep those models at some point.

[VascularHand @VascularHand]

@steeve @hotschmoe @zml_ai thanks, I haven't looked, I had no idea anyone else was working on it.

[VascularHand @VascularHand]

@hotschmoe Been there done that

[VascularHand @VascularHand]

@hotschmoe Is this the vllm docker image from intel? What Linux kernel version are you on?