PaulNg

The hard way is always the right way. Learn from Winners.
People have the POWER.

[PaulNg @GtraderN]

@SUBRATA30016572

[PaulNg @GtraderN]

@EricLDaugh It’s another great day. Thank you Mr President.

[PaulNg @GtraderN]

@DonaldTrumpNat

[PaulNg @GtraderN]

Can we back to days that not always have wild swing premarket…

[PaulNg @GtraderN]

@DrEliDavid 😂😂

[PaulNg @GtraderN]

@OGNodeFather @__tinygrad__ @scaling01 Training a large model may take weeks to months.

[PaulNg @GtraderN]

@jukan05 @AndrewCurran_ Sounds fake.

[PaulNg @GtraderN]

The timing for cut rate maybe far later than people expected. A guess. Because There is a dependent variable.🤣

[PaulNg @GtraderN]

@CharlieVictor62 @whyyoutouzhele 那為甚麼要吹4000億?

[PaulNg @GtraderN]

@JulienTechInvst @JonathanRoss321 Storage is the next silver.

[PaulNg @GtraderN]

@signulll

[PaulNg @GtraderN]

@Samuel85999737 There is a lot of things happening around the world as well. Like wars around the world forced governments sell their gold, bitcoin, etc… to build factories, buy materials. They had places to use money, so the “money” could stay low for a while.

[PaulNg @GtraderN]

It’s lacking of capital now all around, so Gold and Bitcoin are not good for hold.

[h100envy @h100envy]

Ex-Berkeley PhD who leads SGLang at xAI explained how they serve Grok on 100K GPUs in 23 minutes - better than $2000 inference-at-scale courses.

split prefill and decode -> shard experts across GPUs -> route tokens per expert -> overlap comm and compute -> serve at DeepSeek-API-killing prices.

That loop is why xAI runs Grok on SGLang and third parties beat DeepSeek's own API by 5x on cost.

SGLang + prefill-decode disaggregation + expert parallelism + AMD MI300 - that's the stack.

Watch and save it, then read the article below.

[PaulNg @GtraderN]

@TrendSpider

[PaulNg @GtraderN]

@DegenerateNews Great!

[PaulNg @GtraderN]

@saylor 😂

[PaulNg @GtraderN]

@Hedgeye Places like Beijing Shanghai Shenzhen still far above level of 2015.

[PaulNg @GtraderN]

@EricLDaugh 🤣🤣🤣