@HotAisle We talked a ton about storage, slurm/k8s and many other things there. Just read it.
Dylan Patel
653 items collected
653 matching items · newest first
@HotAisle You could read the article and see where you were deficient be other clouds on featureset. Or look at the website Jordan made
ClusterMAX is point in time SOC2 is not the only criteria, but a meaningful one for why you had underperform There are numerous other criteria that your Cloud was deficient in relative to others as explained in the report. We are doing another round of reviews right now for ClusterMAX v2 @JordanNanos reached out already and is going test afaik
@HotAisle You can read all about all the critical missing features and functionality here. It's a legitimate review https://t.co/k51A7YOXMu
@tylerhnorris @CJHandmer @dwarkesh_sp Who says they operate at that high? Everyone is always talking about the impact from the variability of the workload and the spikiness of it
@eric_ruleman Can. You share lots or screen record. I can't repro
@HotAisle These numbers are all nonsense why are you defaming us with bullshit numbers
@eric_ruleman Seems fine to me?
@peckishorigin @bodyforsin Yes lol
@bodyforsin Vibe code a comprehensive scraper for BERC EQR BERC agency EQR is the database I'll pay u $1k and give you a job if you do it by Monday morning. I can get u Claude code access for the weekend if required.
@bodyforsin What can u do
@bilalthreevec How did you not know my ego is absurdly inflated
@typedfemale @cfree94 Wait what. I'm calling you honey?
can I say something without people getting mad?
Are you ready for OpenAI to unleash the monetization slop upon the masses? The real release of GPT-5 was the router and how they can reduce cost / compute for most users by routing to cheaper models. Open has a big strategic shift in the works, and consumer purchasing will shift.
There's a revolutionary change coming to HBM4 with custom base dies Various different accelerators are doing many things differently with custom HBM including OpenAI, Nvidia, and AMD There are many problems beingnsolved including shoreline area, memory controller offload, repeater PHYs, LPDDR + HBM combos, and various beachfront expansion techniques SRAM tags, compute under memory, supply chain implications, and Samsung
I love Gracie Abrams https://t.co/gaSZXD0Q1Q
Apparently the EAs consider eating honey worse than eating chicken which is worse than eating beef because how much animal suffering each brings.
Doesn't matter what I think Here is the incremental user https://t.co/VqS2SREseb
GM America We hold these truths to be self-evident, that all men are created equal, that they are endowed by their Creator with certain unalienable Rights, that among these are Life, Liberty and the pursuit of Happiness. Prudence, indeed, will dictate that Governments long established should not be changed for light and transient causes; and accordingly all experience hath shewn, that mankind are more disposed to suffer, while evils are sufferable, than to right themselves by abolishing the forms to which they are accustomed. That whenever any Form of Government becomes destructive of these ends, it is the Right of the People to alter or to a…
The intern is cooking
Happy to collaborate with High Yield to help him go full time! He's the best small YouTube channel on silicon / semiconductors, so gonna be awesome to see his growth skyrocket
So seems like @yacineMTB wants to rip bong hits. @asianometry are you ready for the 7th 👀 @dwarkesh_sp finally we get to see it https://t.co/TMjKKdzXlp
The intern made a beautiful slop video and threw hella rack pictures for it with great catchy music ... But they forgot to check with anyone on the actual team that the first image is TPU racks lmfao
Prefill on Etched Decode on Cerebras / Groq Would be the most gigabrain inference setup ever and maybe even cost effective
General purpose robots are inevitable They’ll perform any task in any domain Mass labor replacement soon But won’t happen all at once Robots will gain skills level by level Each unlocks capabilities To track progress, we created a Robotics Levels of Autonomy classification system
When you go on vacation you relax When I go on vacation, I go to see the Ferrari that Lisa Su sponsored when she was asked if she speaks English https://t.co/0BCwwNCHqB
i do not understand the huggingface economy. open models are not producing income, how are they buying the gpus? you can say oh they finetune and upload to the hub but who is buying the fine tuned tokens except unemployed users ? where is the liquidity in this ecosystem?
There's a joke to be made here about how Sydney Sweeney is handling a 6" or 8" wafer and not 12" wafer like all leading edge semiconductors are actually made on.
The children yearn to be working in fabs Taiwan high school science exhibition this year are discussing about 1.5nm Gate All Around transistor structure optimization The kids are unbelievably cracked https://t.co/kE8qKCSV4n
Collection coverage
facebook: unchecked
linkedin: linked
podcast: unchecked
reddit: unchecked
substack: linked
web: linked
x: linked
youtube: unchecked