LordOfTheYaml

Dude GPT-Image-2... wtf... how.

CEO Rene Haas on why AI agents are spiking CPU demand: ▫️use >15x more tokens than humans ▫️a 1GW data centres now need 4x CPU cores (120m vs. 30m) for the AI agents AI agents are doing general asynchronous tasks (browse, spreadsheets) that are best-suited CPUs. So, 4x more CPUs per 1GW of GPUs…but the data centre’s power envelope is the same. Need energy-efficient CPU solutions, which is right down Arm’s lane of expertise (its ubiquitous mobile chips win on energy efficiency)….hence ARM’s recent AI CPU announcement.

@poemsekai @Eurostar @eurostarfr @EurostarUK 👍

@Eurostar @Tycale @eurostarfr @EurostarUK thank u!

. @Eurostar @eurostarfr @EurostarUK My train was 4 hours late, and your complaint form cannot be submitted because the CAPTCHA is broken. This is very frustrating. prr.eurostar.com/?lang=fr#/PRRR… #Eurostar

@EurostarUK @Eurostar @eurostarfr Thanks for the fast resolution! ☺️

@Tycale @Eurostar @eurostarfr Hello, We're really sorry to hear about the delay and the online issue. Could you please share your booking reference so we can look into this for you and help you move forward with your claim? Thank you for your patience. twitter.com/messages/compo…

to be clear, NVIDIA is NOT a car

@GLBouchez @IENouwen Être libéral signifie que l'on souhaite réellement diminuer les responsabilités de l'État de manière durable. Cela implique que l'on préfère éviter de créer de nouvelles dépenses ou de nouveaux manques à gagner. Car ces dépenses et ces manques à gagner impactent la dette belge.

@IENouwen Je suis libéral de droite: je défends ceux qui travaillent

Nog zo’n poseur die pretendeert rechts-liberaal te zijn.

my clanker trying to build data intensive applications

Je pense que quasi personne ne connaît ces chiffres : notre société prend collectivement en charge 86% des dépenses de santé et 95% de la vieillesse. Et seulement 26% des coûts liés à la natalité. Nous vivons donc dans une société qui néglige la seule fonction indispensable à sa survie : la reproduction. C’est pour cela que cela me rend fou quand j’entends que vraiment on ne sait pas quoi faire contre l’effondrement démographique, que c’est une espèce de fatalité. Non. Nous mobilisons chaque année de plus en plus de ressources collectives, laissant donc de moins en moins de ressources personnelles, et nous les consacrons de plus en plus à autre chose qu’à la natalité. Ainsi chaque année les jeunes couples ont un peu moins de ressources disponibles pour faire des enfants. Puis nous nous étonnons que la natalité s’effondre. Quelle folie.

@TheInsiderPaper Mythos is currently in pursuit of new sets of training data.

A hacker has allegedly stolen a vast trove of sensitive data—including classified defense documents and missile schematics—from a Chinese state-run supercomputer, in what could be the largest data breach in the country’s history. The dataset, reportedly exceeding 10 petabytes, is believed to have been taken from the National Supercomputing Center in Tianjin, which serves over 6,000 clients, including scientific and defense institutions — CNN

We've signed an agreement with Google and Broadcom for multiple gigawatts of next-generation TPU capacity, coming online starting in 2027, to train and serve frontier Claude models.

@d29495429 @realgabe178 @jessegenet @Kaihan1984 @GoogleDeepMind @openclaw Sad truth, yes. Still, it's a lot of fun ! 😄

@Tycale @realgabe178 @jessegenet @Kaihan1984 @GoogleDeepMind @openclaw Totally agree and I’d go further tbh local models are just for messing around. Waaaay more cost effective to use API for almost all use cases

It’s happened. Mac Studio is here. Gemma 4 31b @GoogleDeepMind installed, chatting with my main @openclaw for $0 in token expenses now... I've burned $5-6k on tokens on my crazy ideas over past few months, so this mac studio should pencil out for me within 3 months or so 🤓

@realgabe178 @jessegenet @Kaihan1984 @GoogleDeepMind @openclaw Make no sense to buy that much memory.. Nvidia is king to run 31b. RTX 5090 32gb fires 45tok/s with 65k tok window.

@jessegenet @Kaihan1984 @GoogleDeepMind @openclaw To run a 31b model 😳

if you don't have these in your configs you're ngmi

My dear front-end developers (and anyone who’s interested in the future of interfaces): I have crawled through depths of hell to bring you, for the foreseeable years, one of the more important foundational pieces of UI engineering (if not in implementation then certainly at least in concept): Fast, accurate and comprehensive userland text measurement algorithm in pure TypeScript, usable for laying out entire web pages without CSS, bypassing DOM measurements and reflow

Wtf did they do to Overpass?????

@0xSero github.com/aliez-ren/vllm… with nvfp4 and vllm

Got a 5090 for a few hours, got: Qwen3.5-27B-AWQ-4bit with 81920 context up from a max of around 44688~ with vllm github.com/0xSero/turboqu… I don't trust any of these numbers 100% so people poke around if u can test this out

@vikhyatk @snwy_me :O

@snwy_me idling my node just to feel something

where the fuck are all of the GPUs going?!? i need literally one 8xH100 node and i cannot for the life of me get one ANYWHERE

Someone made a game about SCRUM meetings destroying productivity.

For OpenClaw, just use Qwen3.5 27B! Q4 GGUFs match the original's model accuracy You don't need expensive hardware or models