The Ledger, every show

“LLMs are only going to carry us so far into AI, but all of these labs are not going to find product market fit and business value. Like it's just not going to happen.”

“There's no hardware that you can't hack with one of these things.”

“Everyone that I've been interacting with and many people across the open clock community are switching it out for Opus four dot eight and they don't notice a difference.”

“So like, what does that mean for all of these enormous positions that are out there in these frontier labs? I think it's kind of shaky.”

“I'm hearing that they can reverse engineer any software binary. And so like software is no longer open Closed source. All software will be open source going forward because you can just rip open any binary and reconstruct the software.”

“And then he hired forensic experts to go and factually prove that it is China actually funding all of this opposition in Utah. And he actually mapped it a lot. You can go look on his Twitter and Instagram and he shares all of the data. And apparently has now s…”

“I think this is, there's like a likelihood that the inequality that has been driven by software entrepreneurship up until this point is that kind of a local maxima and that because today anyone in the world can create software at a pace that was impossible six…”

“Well, what I'm actually hearing is that they're trying to make open source illegal at the federal level.”

“You've got big foundation model companies are out there buying the slack instances of failed startups to get more data.”

“Anthropic says no one will be doing software engineering 12 months from now.”

“Now I will say one, but saying the other thing on the diff thing I'm hearing is that nobody's using Git anymore. They have these like merge bosses that, you know, basically”

“Yeah, just what I'm saying is that like, if you're investing in something now or you're an entrepreneur starting something like you're what you're shooting a rocket that lands five years from now. Right. And so five years from now, like this is going to be cra…”

“And I think we're sort of rapidly moving towards a world where the UI, which largely was the entire business model of SAS for the last 10 years, most of SAS was like, I'm going to build this complex UX on top of a Postgres database, then I'm going to sell it t…”

“Whatever the compute configuration plus the software configuration that was going on in Q two of 20, 24 at open AI was the most powerful, most awesome version. And everything's been downhill from here.”

“Yeah, I mean, that goes back to that first research paper that came out right after ChatGPT launched, which basically said you can't create Fake data for training models. It just doesn't work, and so there was, like, this very early paper that effectively told…”

“You and I both know that we had a founder Who's from Australia that went to renew his visa and had to plug his phone into a American computer with the I don't know which department, I guess it would be the state department. His phone, all the content in his ph…”

“If you've got an incremental dollar, putting it into the S into an S and P 500 index fund for the most part performs better than almost all of the venture capital asset class.”

“I think a year from now you know, most programming will be done in English. And I think that's the, there's almost no question.”

“We were meeting with a very large LP today who summarily said to this we have a lot of data on this and we can definitively say that the venture capitalists that are tweeting the most are not the ones producing the DPI.”

“Tesla has also assembled the largest supercomputer in the history of humanity. That's a Tesla asset that is the GPU cluster that they've put together.”

“A simple way to think of it is like, we're probably gonna be kind of where we are right now for a lot longer than people think. If the current architecture and the current energy and the current capital needs are the gating factors.”

“One, the value of all software being developed is kind of rapidly going to zero right now because anyone can build anything that they want for any workflow in their business or their personal life very, very quickly and for very low cost. And so that means tha…”

“There's actually a whole lot of people involved in these things that are not actually students. And I think that really confuses things and almost makes these things like deep fake protests.”

“Here's a bunch of teams, not just one, but many AI teams standing on the precipice of making a moral decision. And they knew they either needed to pay for it or steal it. And some of the teams chose to steal it instead of paying for it.”

“Like, there's no way that doesn't happen over time. It's too appetizing. The data asset is too, Like, the centralization of that much data, of that much value, like, you cannot keep your hands off of it.”

“It's kind of like Meta trying to get into hardware. Like they're just never going to be good at hardware.”

“They will be in the public markets by 27.”

“I think this is actually the beginning of Grok being the third foundation lab that is truly in the game.”

“I think you're going to see a retraction in the market. Like you're going to see we've built out a dramatic amount of infrastructure. A huge, you know, three trillion, four trillion dollars of capex has gone into this, and everyone rushed in, and they're tryin…”

“Both the frontier models from the American companies have specific changes that have been made in their harnesses that hobble the model's ability to respond to certain types of tasks. And in particular, the two types of tasks that are the most on alert on watc…”

“The truth is there's about, you know, there's less than ten million people using these things.”

“All of these frontier model companies are unprofitable on training, right? It costs an enormous amount to train each subsequent generation. They appear to be profitable on an inference basis.”

“And the open source models are less than two months behind.”

“So, you know, Mira Mirati's thinking machines appears to be a really solid Open weights model. You know, it's a almost nine hundred and seventy five billion parameter model million token context window mixture of experts. It's really good. And this thing's ope…”

“And I just think that that's the simple thing here, which is that AI is less popular than ICE and Jeffrey Epstein.”

“You know, I think if you look at NVIDIA, you know, rent, spot rental on NVIDIA servers is down, like, 22% week over week the last couple of weeks. But all of the capital is shifting to, Buying actual raw metal.”

“And we got this thing up and running and out of the box, it's doing a 150 tokens per second, which is like 10 times what you get out of frontier models.”

“And what it looks like they've done here is just taken, you know, state of the art Gemini models, both the on-device local flavor, as well as the cloud flavors and distilled them for their own purposes into these, you know, quote, Apple foundation models. Whic…”

“SpaceX reserved up to five percent of the IPO shares for a directed share program, which is people selected by the executive officers. And so they're offered at the IPO price and they have no post IPO lockup restrictions.”

“They've also got a Desktop that can do a one trillion parameter model, which is pretty amazing. We had it on stage with us.”

“The soundbite is that there are AI companies, I'm not, doesn't matter which ones, that are hiring the Former employees of ad agencies to do their job, to play act their old job, and then recording all of it in order to get the training data to learn marketing …”

“ChatGPT itself still isn't beyond eight hundred million people, and it's declining in growth right now.”

“Every entrepreneur in my portfolio is like swapping out open source models for everything. And of course you would. Right. Cause that's your margin.”

“And so you have this thing where open source is also always going to be lagging the frontier. And so it's not going to be like a total replacement.”

“Every single model company is out of data and they need more human generated data in order to make and improve their models.”

“The through line on this is that people are going to be able to request the software that they want in like one to two years from now, and it will just materialize onto the internet.”

“The sheer level of, I mean, one trillion dollars of demand for their chips over the next year.”

“We've got a lot of people from China, you know, Baidu Xiaomi, like many, many different companies that are all developing. Some are developing custom models just for open claw. Some are integrating into their clouds.”