AkhenOsiris
Chanos has entered the chat! 😂
tweet
Chanos has entered the chat! 😂
One of the great myths of the Dot.com Bubble was that Internet traffic was doubling every three months. It was still growing 100% every year, but not 1,000%. This miscalculation had huge implications for telecom capex going forward. - James Chanostweet
AkhenOsiris
Zuck: My earnings call is before yours Jensen! This DeepShit Chinese bullshit is overshadowing a good qtr.
Jensen: Ok, hear me out...Llama is open source...so is DeepShit...and it's doing some incredible stuff with much lower cost, so tell them you are going to copy their shit and it is extremely bullish for Meta, as FCF gonna ramp like a motherfucker.
Zuck: That is fucking awesome, ok, I'm going to pound FCF and cost discipline down their damn throats. But one thing...that means I am buying less of your GPUs no?
Jensen: Yes, I am fucked! But I have a few more weeks to figure out how to sell our inferencing capabilities and how Agents and Physical AI like robots are basically here (even though they aren't).
tweet
Zuck: My earnings call is before yours Jensen! This DeepShit Chinese bullshit is overshadowing a good qtr.
Jensen: Ok, hear me out...Llama is open source...so is DeepShit...and it's doing some incredible stuff with much lower cost, so tell them you are going to copy their shit and it is extremely bullish for Meta, as FCF gonna ramp like a motherfucker.
Zuck: That is fucking awesome, ok, I'm going to pound FCF and cost discipline down their damn throats. But one thing...that means I am buying less of your GPUs no?
Jensen: Yes, I am fucked! But I have a few more weeks to figure out how to sell our inferencing capabilities and how Agents and Physical AI like robots are basically here (even though they aren't).
tweet
AkhenOsiris
Some of the hilarity with this headline:
- Government cannot be happy with this
- Unlike TikTok, nothing to ban here, open source and all over the world now
- Furthermore, banning isn't the answer anyways. Copying is.
tweet
Some of the hilarity with this headline:
- Government cannot be happy with this
- Unlike TikTok, nothing to ban here, open source and all over the world now
- Furthermore, banning isn't the answer anyways. Copying is.
JUST IN: Chinese AI platform DeepSeek overtakes ChatGPT on Apple's app download rankings in the US - The Spectator Indextweet
AkhenOsiris
My dad is confused, "why is it bad for the market that DeepSeek is making it cheaper to run AI?"
I asked him if he understood why certain stocks go down when "Amazon your margin is my opportunity" enters a market. He said of course I do.
The tantrum tonight is about margins for semi infra. Warranted? Maybe. If compute explodes (as you'd expect), maybe not? Maybe NVDA GM has peaked?
Regardless, markets always shoot first and rationality returns later. The fact that megacap earnings calls start this week, hopefully provides some clarity.
tweet
My dad is confused, "why is it bad for the market that DeepSeek is making it cheaper to run AI?"
I asked him if he understood why certain stocks go down when "Amazon your margin is my opportunity" enters a market. He said of course I do.
The tantrum tonight is about margins for semi infra. Warranted? Maybe. If compute explodes (as you'd expect), maybe not? Maybe NVDA GM has peaked?
Regardless, markets always shoot first and rationality returns later. The fact that megacap earnings calls start this week, hopefully provides some clarity.
tweet
AkhenOsiris
RT @TShirtnJeans2: Replying to @_fabknowledge_: Pre DeepSeek @kakashiii111 o3 cost $3000 a run, will bankrupt every AI startup and NVDA will crash.
DeepSeek releases R1 @kakashiii111 DeepSeek made AI so efficient to run on Nvidia GPUs startups will never need to buy more and NVDA will crash.
tweet
RT @TShirtnJeans2: Replying to @_fabknowledge_: Pre DeepSeek @kakashiii111 o3 cost $3000 a run, will bankrupt every AI startup and NVDA will crash.
DeepSeek releases R1 @kakashiii111 DeepSeek made AI so efficient to run on Nvidia GPUs startups will never need to buy more and NVDA will crash.
tweet
Offshore
Photo
Hidden Value Gems
Interesting if this has spooked the markets this morning 📉
High expectations, billions of capex, high pace of innovation. Tempting to dive in but hard to have an edge as an investor...🎲
#stocks #AI https://t.co/tiVsgibFxk
tweet
Interesting if this has spooked the markets this morning 📉
High expectations, billions of capex, high pace of innovation. Tempting to dive in but hard to have an edge as an investor...🎲
#stocks #AI https://t.co/tiVsgibFxk
tweet
Offshore
Photo
Hidden Value Gems
"Many people are speculating that DeepSeek is simply lying about the number of GPUs and GPU hours spent training these models because they actually possess far more H100s than they are supposed to have given the export restrictions on these cards, and they don't want to cause trouble for themselves or hurt their chances of acquiring more of these cards.
While it's certainly possible, I think it's more likely that they are telling the truth, and that they have simply been able to achieve these incredible results by being extremely clever and creative in their approach to training and inference. They explain how they are doing things, and I suspect that it's only a matter of time before their results are widely replicated and confirmed by other researchers at various other labs."
h/t @doodlestein
Good read 👇
https://t.co/dWEKGCHlUH
#NVDA #TSMC #ASML #META #GOOG #DeepSeek
tweet
"Many people are speculating that DeepSeek is simply lying about the number of GPUs and GPU hours spent training these models because they actually possess far more H100s than they are supposed to have given the export restrictions on these cards, and they don't want to cause trouble for themselves or hurt their chances of acquiring more of these cards.
While it's certainly possible, I think it's more likely that they are telling the truth, and that they have simply been able to achieve these incredible results by being extremely clever and creative in their approach to training and inference. They explain how they are doing things, and I suspect that it's only a matter of time before their results are widely replicated and confirmed by other researchers at various other labs."
h/t @doodlestein
Good read 👇
https://t.co/dWEKGCHlUH
#NVDA #TSMC #ASML #META #GOOG #DeepSeek
tweet
Offshore
Photo
InsideArbitrage
https://t.co/ke4keehAAu $JD Proposes to Acquire Dada Nexus $DADA -
✴️https://t.co/ke4keehAAu proposed to acquire all outstanding ordinary shares of Dada at a purchase price of $2.00 per ADS, or $0.50 per Ordinary Share, payable in cash.
✴️The Board has established a Special Committee composed of three independent directors: Baohong Sun, Laura Marie Butler, and Jian Han. Laura Marie Butler will serve as the chair of the Special Committee.
tweet
https://t.co/ke4keehAAu $JD Proposes to Acquire Dada Nexus $DADA -
✴️https://t.co/ke4keehAAu proposed to acquire all outstanding ordinary shares of Dada at a purchase price of $2.00 per ADS, or $0.50 per Ordinary Share, payable in cash.
✴️The Board has established a Special Committee composed of three independent directors: Baohong Sun, Laura Marie Butler, and Jian Han. Laura Marie Butler will serve as the chair of the Special Committee.
tweet
Offshore
Video
Startup Archive
Alex Schultz on how Mark Zuckerberg cut the “gordian knot” of Facebook’s analysis paralysis
At the end of 2007, Facebook’s growth began to stall, which prompted them to create their famous growth team.
One of the first things the growth team did was try to figure out Facebook’s magic moment. But as Alex explains, arriving at their famous “7 friends in 10 days” magic moment wasn’t straightforward:
“We had this huge discussion internally because we saw a strong correlation between number of friends and percent active. But there was this huge conversation internally about whether it was causal or whether it was correlated…. We had this huge fight going on between data science, data analytics, engineering, product, and marketing, and we couldn’t get to a decision.”
Then Mark Zuckerberg just made a decision.
“Mark simplified things down and said, ‘Get everyone to 10 friends in 14 days.’”
And when the team did this, retention improved dramatically.
“It turns out making that change happen was the thing that told us whether or not it mattered… Cutting the Gordian Knot and making the decision to move forward was the most valuable thing that Mark did for the growth team because it took the shackles off of us and allowed us to just say, fine, we’re going to get people connected to their friends. And that was the magic moment for Facebook.”
Alex believes there’s an important lesson here for all startups:
“With every company that I have advised or worked with, there is a magic moment. There is a thing about your product that people love and their eyes light up. You often find that qualitatively, and then you back it up with correlations in the data. But it’s very hard to prove causally — especially when you’re very small and don’t have a lot of data. People will argue and argue about it. But the best way to find out if they’re right is to break it or make it move up tremendously and then see what happens.”
Video source: @khoslaventures (2015)
tweet
Alex Schultz on how Mark Zuckerberg cut the “gordian knot” of Facebook’s analysis paralysis
At the end of 2007, Facebook’s growth began to stall, which prompted them to create their famous growth team.
One of the first things the growth team did was try to figure out Facebook’s magic moment. But as Alex explains, arriving at their famous “7 friends in 10 days” magic moment wasn’t straightforward:
“We had this huge discussion internally because we saw a strong correlation between number of friends and percent active. But there was this huge conversation internally about whether it was causal or whether it was correlated…. We had this huge fight going on between data science, data analytics, engineering, product, and marketing, and we couldn’t get to a decision.”
Then Mark Zuckerberg just made a decision.
“Mark simplified things down and said, ‘Get everyone to 10 friends in 14 days.’”
And when the team did this, retention improved dramatically.
“It turns out making that change happen was the thing that told us whether or not it mattered… Cutting the Gordian Knot and making the decision to move forward was the most valuable thing that Mark did for the growth team because it took the shackles off of us and allowed us to just say, fine, we’re going to get people connected to their friends. And that was the magic moment for Facebook.”
Alex believes there’s an important lesson here for all startups:
“With every company that I have advised or worked with, there is a magic moment. There is a thing about your product that people love and their eyes light up. You often find that qualitatively, and then you back it up with correlations in the data. But it’s very hard to prove causally — especially when you’re very small and don’t have a lot of data. People will argue and argue about it. But the best way to find out if they’re right is to break it or make it move up tremendously and then see what happens.”
Video source: @khoslaventures (2015)
tweet