RT @mcpburner: Help me find all my friends 🥂

@fan_of_y3 lol

RT @noetic_emetic: me when i make a priori synthetic valid judgements

@fan_of_y3 We are reading city of god

@rustbeltkid1 @MaxWithNanos Repeal ethanol subsidies. As someone from the cornbelt, I want my natural environment back and free of pesticides.

RT @badazn: ye is a filter those who get offended by what he posts show low impulse control/low iq & ultimately hate freedom with their de…

@arpalargodesa @jlippincott_ “You couldn’t not pay me to marry outside my ethnicity” I’m with Marko Elez

RT @fan_of_y3: The real red pill is that the country essentially ran itself during the Biden years, the infinite bureaucracy doesn't need l…

RT @cryptolyxe: We're gonna get raped so much you may even get tired of getting raped and you'll say please please it's too much raping we…

@SkierGroyper45 Donda gets so much hate, but it’s top 5 Ye

@indeliblemarker @Logo_Daedalus No smart person expected the 2022 export restrictions to cripple China. At that time China could already produce many of the chips that were banned for export, and they are still relying on western lithography equipment and EDM software for their advanced processes.

@indeliblemarker @Logo_Daedalus He says: “Well, obviously China has gone way ahead.” What is he referring to that China has gone way ahead in then?

@ShoneSanchez @Hogheaded11 @Logo_Daedalus I don’t think you understand open source, there is no ‘closing the doors’

@ShoneSanchez @Hogheaded11 @Logo_Daedalus Meta has always open sourced their SOTA llama models, so not even that big of a development.

Deepseek had some great innovations in post-training RL, which made training process far more efficient. However, given the rate of progress in LLM development, Deepseek’s model is about in-line with expectations of what a model released now should do. The large jump is because OpenAI, Anthropic, Meta, etc. have slowed down their product releases and focused on internal research. What’s significant is that it is a Chinese company at the state of the art.

@Hogheaded11 @Logo_Daedalus 92% of REVENUE, nowhere close lol

@alessio_joseph And it’s beautiful

@Logo_Daedalus 108B is about 12% of their revenue over that period, less than most tech companies typically do now. Blaming stock buybacks is misdirection for bad management and technical issues internally. Also he’s wrong about Chinese being ‘ahead’ of US semi tech, but that’s another thread.

@Logo_Daedalus The avg company spends ~5% of revenue on buybacks a year, and it’s true that QCOM spends more than that, ~20%. Goalpost shift though, Hudson was wrong about Intel.