mukund

ai engineer & programmer. code, research papers, random thoughts 𓇼

[mukund @thwmjha]

@leviackermanft nice bro!

[mukund @thwmjha]

@russoh121 @PhysInHistory man, you look like a talented physist, let's connect.

[mukund @thwmjha]

@soham_btw huang gone, nvidia domination gone..

[mukund @thwmjha]

@ArjanChaudharyy cool stuff bro!

[mukund @thwmjha]

@fj45y @ParamSiddh it might be better buddy,but i mean in my mind, gemini is a general focused llm, not coding llm.

[mukund @thwmjha]

@strwbcom they should not have done this thing to such loyal person who built them brick by brick. btw nvidia downfall starts now.

[mukund @thwmjha]

@fawnunwrap true. if they try, they can do good in new inventions as well, especially deepseek.

[mukund @thwmjha]

@wavefnx ship-fast turns products into performance. only time proves integrity.

[mukund @thwmjha]

@fawnunwrap let me push u, but push me back as well.

[mukund @thwmjha]

@kartikktwt bhai nokia ka bhi llms.text hai. Wtf.

https://t.co/d5ZfNvbZsT

[mukund @thwmjha]

@facedevstuff @SimOnX_PL @OpenAI not arguing, but windows should not be used by people like u who do programming.

[mukund @thwmjha]

@ArjanChaudharyy same to u bro!

[mukund @thwmjha]

@vikhyatk i teach computers @moondreamai

[mukund @thwmjha]

@ParamSiddh Yupp.

[mukund @thwmjha]

the kv cache decides your batch size. pagedattention borrows os paging to waste almost none of it, for 2-4x throughput

[mukund @thwmjha]

@leothecurious Datacenter 9/11

[mukund @thwmjha]

@strwbcom using shears to cut your hair.