SHORT BIZD

“About three weeks ago, I had a problem with my with my VPN on my computer. I was using Claude to kind of help me through figuring out how to get it fixed. I'm a software developer by background. I have an engineering degree from Stanford. You know, I I I do know how to program quite well, but I kinda veered into PowerShell commands in Windows 11, which I don't know very well. And I was blindly pasting in Claude, and it made me delete my entire networking stack. And, you know, so you have these very crazy failure modes with AI where you either go in loops or it goes away in thinking and and, you know, it goes and, you know, burns, you know, 10 x the inference”

“And it gets even worse on these $200 plans, which are used by programmers now because the nature of programming is you're streaming tokens almost forever. And, you know, the average user on a, on a $200 plan can burn, you know, many thousands of dollars of underlying compute. And in fact, there's a there's a leaderboard called Vibrank, which, where people compete to see how much money how how much underlying compute they can burn on a $200 plan, and the leading guy on the, on the leaderboard has burned 51,000 US dollars in a single month on a $200 plan.”

“China got the message out of the stories that have been published in major news media outlets and very prestigious outlets. We've seen a number of stories talking about Iran closing the Hormuz Strait. Again, we are talking about June. So what I wrote at that time was China got the message, and the message was the homeless rate will be closed. But it's not Iran who is going to close the Hormuz Strait. It is The United States. We've been talking about how China was building those massive inventories, oil inventories and everything else, in preparation for either war, sanctions, or a major interruption. Now we know. China knew.”

“If this is going to drag on until May, our modeling shows that the whole globe, basically, is going to suffer a major recession and probably stagflation. When you talk to economists about the current, about the impact of the current crisis, the difference between them both of them both views see a recession, but they disagree on inflation. So are we going to have stagflation or just a recession? Regardless, this global recession basically is going to reduce global demand significantly to the level where prices will decline. So, yes, we see prices going up until we hit that recession, and then prices will decline again.”

“While you you are riding on that sort of silicon Moore's law trend, because the models are in such a brutally competitive environment where there's, literally zero lock in, there's, you know, almost zero switching costs so I can and, yeah, if one day chat GPT five comes out, I'll switch to that, but then cord 4.6 opus comes out and that's better. I'll switch to that, and so there's nothing stopping me overnight really just changing the models. What the situation basically is that each generation of model is being rushed out to kind of get the top of the scoreboard so that all the the customers flock to that. And as a result of that, the actual amount of inference or, tokens burned, if you will, per useful query with each generation of model while the underlying compute is getting cheaper and cheaper in terms of, what you can get done on the chips, the amount of inference you have to burn for a useful query is actually going up quite dramatically.”

“Even if you send the navy to escort ships out of the Hormuz, ships are not going to move without insurance. If the insurance remains nonexistence or extremely expensive, ships are not going to go along with the navy because shippers are going to tell Trump or Modi or anyone else who wants to send those navy vessels, telling them, look. If I go with you and I get attacked and I lose my ship and I lose everything on it, will you compensate me? And Trump is going to tell them no. They need that insurance to exist and to be cheap enough for them to move. So anyway you look at it, you say it's conspiracy theory or not, it doesn't matter. They need that insurance. So it is an insurance story anyway you look at it.”

“AI is a power tool, and so is a chainsaw. Right? You give a chainsaw to a carpenter, and they can do a skilled carpenter, and they do amazing things. You give a chainsaw to a novice, and you can cause all sorts of all sorts of problems. And so what I'm observing, both across my platform as well as within my my engineering team is the the people who are benefiting most from AI, the ones that are highly skilled and intelligent and know how to use the AI, and they're seeing productivity gains, which are astronomical. They're seeing, you know, double to triple their productivity”

“I think the interesting thing here is that it won't be the government stepping in. I think that there's one big company that's kind of been sitting in the wings that has a god awful AI product, and, it's in everyone's pockets, and that's Apple and the iPhone. As with the .com bust, there was a lot of infrastructure that went through to the second and third owners where, you know, maybe the first owner that builds out the optical fiber, network goes bust and the second owner comes in and tries makes it to be breakeven after after, you know, purchasing it at a cheaper price and washing out the underlying sunk costs, and then the third owner comes in and and makes money out of it. And I think you might have that situation in the AI compute space”

“The headline number is 122,000,000,000 raised on a 730,000,000,000 pre money valuation. When you get down to the actual segmentation of kind of what's going on, it's it it seems that it's only about $25,000,000,000 worth of cash. It's and it seems to be more of a vendor financing than an actual straight cash injection. You've got Amazon, Nvidia, and, SoftBank primarily putting the money in or the in kind in. Amazon's putting in 50,000,000,000, but that's contingent on OpenAI spending a 100,000,000,000, I think, over the next eight years on their compute”

“It really turns software development into a slot machine in that, if the if if, ultimately, you know, you you you know, you're writing your app and you you kind of need to get get caught to do something rather, you're really pulling the slot machine handle. You don't know how much gonna how much it's gonna cost you by the time the tokens are all burned, and you don't know if you're actually gonna get a a solution at the end of the day. And so, you know, I mean, I guess, only in Silicon Valley could could they turn software development into to generate gambling because that's that's kind of where you end up. And it will look at you in the eye with the eyes of a sociopath on a first date in some regards trying to, you know, gaslight you into thinking that its answer is is true when it's it's clearly not the case.”

“Well, this is where the fifth industrial revolution meets the Islamic revolution. Right? A lot of the financing for AI has come from The Middle East, and you can imagine now if you're Saudi Arabia or you're The UAE and you have a fiduciary duty to protect your nation and your citizens and your economy, are you gonna be putting it into a hyper rounds of Sam Altman's, highly inflated valuations, or will you spend on defense energy rebuilding rebuilding your your civilian and and industrial infrastructure and potentially going to war to to fund an army?”

“Major nuclear accidents like Chernobyl when they have happened in the past as accidents have set the nuclear power industry back by 15 when they happen. The public gets irrational and fears that a nuclear accident is somehow going to cause the nuclear plant to blow up like a nuclear bomb. So an intentional breach of containment of an operating nuclear power reactor, like the ones at Basher in Iran or Baraka in The UAE, could cause a contamination event an order of magnitude worse than Chernobyl. No military in the history of, the human race has ever intentionally targeted a operating nuclear reactor core and tried to breach its containment intentionally as an act of war.”