Project Star Gate - The Crash of AI and Tech

[John Badman]

I have it on good account that things are in lalaland and everyone is pumping their bags before the sell-off copy trading begins. Market sentiment is poorer than expected despite the soft numbers indicating J Powell is furthering rate cuts on Wednesday. The issue isn’t the market but the overhead of Tech holding a dark money anvil that’s about to drop like a rock on our head. Everyone wants AGI and for crypto to be treated as a security including Larry Fink and Davos.

This is the biggest top signal I’ve seen…

I’ve heard through the grapevine that GB200 rack orders are being canceled because the diminishing returns aren’t seen as marginally worth the energy expenditure cost nor the new infrastructural development. SMRs are the linchpin holding this trillion dollar expedition together but nuclear reactors in general are incredibly finicky to get into production considering location and regulation.

These 30x days are not long for our reality. At a certain point I can healthily admit in realization that my portfolio is completely detached from the real economy. Regardless of the massive profits I’ve made the last few months riding OKLO and NVDA, I’m thinking about taking it all off the table after seeing the writing on the wall. I feel like I’d be better being the first one out the door rather than being margin called next month.

Everyone, and I mean everyone, is thinking about Recession right now. It’s a matter of when.

 

[Kaspersky]

Well, if SoftBank is in on trying to get a company to $500B valuation that's the time to leave.

 

[WeWuzFinns]

Launch niggers into space so that aliens can wipe them out

 

[Tablet County]

Remote Viewers would be cooler.

 

[John Badman]

It begins

 

[Audit]

In other news, having a stable job unrelated to the AI bubble that's about to burst is probably the best move to make right now. Beware the flood of ex-AI "developers" infesting the rest of the tech landscape when companies cut back on their AI investment.

Intel better finally use this as a chance to shed every single jeet on staff and hire no one to replace them in their already immensely bloated pool of staff.

 

[Fascist Kirk]

These Chinks made an ai model for 5 percent cost what all the other AI's Services did. Proof being first is not best, just the most expensive. A million Tech Hodlr's Cried out in terror and went silent.

I new the AI bubble was bullshit, when people are dumping money into server racks, NVDIA chips that overheat, its all one big pump and dump. and today has been great to watch.

Good thread you Called it @John Badman

 

[NoReturn]

In other news, having a stable job unrelated to the AI bubble that's about to burst is probably the best move to make right now.

Where to? Even construction seems iffy right now. People can't afford homes and even with RTO everyone and their mother knows the mandates are just soft-layoffs from commercial real estate renters.

 

[macrodegenerate]

These Chinks made an ai model for 5 percent cost what all the other AI's Services did. Proof being first is not best, just the most expensive. A million Tech Hodlr's Cried out in terror and went silent.

I new the AI bubble was bullshit, when people are dumping money into server racks, NVDIA chips that overheat, its all one big pump and dump. and today has been great to watch.

Good thread you Called it @John Badman

Deepseek has been a sleeper for a while, but its latest models before R1 were mostly designed for programming. I would even call Deep Seek Coding V3 State of the art for coding. The issue is the lite version was beat out by Gemma 2 or Mistral Code, so nobody really played much with it even though the pricing was absolutely insanely cheap for the larger version. Plus the Local LLM community is pretty much all gooners obsessed with ERP. A coding LLM isn't useful for that.


Nobody expected this because nobody was paying close enough attention to how low pricing was on Deepseek. If you have nothing to recoup because your resources are used more efficiently, your pricing will reflect that. If your product is cheaper while being as good or better than the competitor, you've won. Full stop.

 

[Fascist Kirk]

Deepseek has been a sleeper for a while, but its latest models before R1 were mostly designed for programming. I would even call Deep Seek Coding V3 State of the art for coding. The issue is the lite version was beat out by Gemma 2 or Mistral Code, so nobody really played much with it even though the pricing was absolutely insanely cheap for the larger version. Plus the Local LLM community is pretty much all gooners obsessed with ERP. A coding LLM isn't useful for that.


Nobody expected this because nobody was paying close enough attention to how low pricing was on Deepseek. If you have nothing to recoup because your resources are used more efficiently, your pricing will reflect that. If your product is cheaper while being as good or better than the competitor, you've won. Full stop.

its an open source ap what Open Ai was supposed to be. it is a spy bot for the Chinese government, but it does everything the AI boom in mag7 blew into a huge bubble, for 1/7th the cost. the technology does not need power plants server centers, and trillions in investment, it needs better coders.

Off topic but in the sphere wtf Meta earnings? 45 minutes late cut all 2025 guidance .... explains how every yt video I click on wants me to get a meta account .....

 

[macrodegenerate]

its an open source ap what Open Ai was supposed to be. it is a spy bot for the Chinese government, but it does everything the AI boom in mag7 blew into a huge bubble, for 1/7th the cost. the technology does not need power plants server centers, and trillions in investment, it needs better coders.

Off topic but in the sphere wtf Meta earnings? 45 minutes late cut all 2025 guidance .... explains how every yt video I click on wants me to get a meta account .....

If you have a NVIDIA DGX B200 sitting in your basement, you can run it without an internet connection. The Chinese spyware shit is all FUD unless you are actually sending it to their servers. You don't have to, and I promise you American companies will start running R1 on their servers and start offering it as a service. Then you can pay Americans companies to violate your privacy instead.


Something most people aren't talking about is the distilled models. Essentially after making R1, they decided to release it for smaller hardware. The issue is that they haven't trained a smaller version yet. What they decided to do was distill all of the popular 70B and smaller models using r1. Every distilled version has seen significant improvements over the base. If Deepseek releases a 2B / 8B / 60B model of pure R1, it's over. They already have with Deepseek v2, so it's only a matter of time.

 

[General Emílio Médici]

OpenAI literally never made money and since going closed source it just rode a massive hype wave where they would hype up the next big thing that was about to change the entire world, a bunch of marketing would be done, and the product released eventually shown to be a marginal improvement at best or nothingburger at worst.

I read some decent articles on it, explaining not only how unsustainable the whole thing is ( https://www.wheresyoured.at/subprimeai/ ) but also giving a sound theory of WHY there is so much shilling on AI going on in the market ( https://www.wheresyoured.at/saaspocalypse-now/ Tl;Dr big tech wants AI to justify forcing people into buying even more Software-as-a-Service shit like Teams and Slack to continue perpetual growth.)

I can see why NVDA would collapse, but why would OKLO?

 

[NoReturn]

OpenAI has this form on their website.

https://openai.com/form/stargate-infrastructure/

How many responses do you figure it has so far?

 

[Zbk3VmLfujjzjxBd43tL]

This thread should join the FUD threads from 2022, 2023, and 2024 about the imminent crash of the market. I'm putting "people on Kiwi Farms are saying the market's about to crash" on my list of "buy buy buy" signals since you guys are about as reliable as Jim Cramer.