LLMs

“There's a certain kind of a honeymoon period happening with GPT and LLMs, which I think will find the right limits to it.”

“It's how it will work, of course. You have, you know, every two weeks, incredible things coming out of research, and if you just assume it's a static, in a year from now, your product will be very dumb compared to your competitor's products. Or more expensive …”

“I've heard there are several companies now that have grown rapidly in ARR and have now started declining in ARR as retention actually starts to affect these businesses because there's churn from folks who just try out products built on top of these LLMs becaus…”

“In the large language model world, that the LLMs themselves will serve as a use case versus an application layer on top of those LLMs.”

“The reason is nobody could predict that LLMs would be so useful and powerful before you try to run at this scale.”

“I think as a function of LLMs, trust centralizes, and there's huge economies of scale to subscription businesses and advantages, and so it's a silly, silly fun idea, but if we're in media.”

“What you're seeing many of us, Snowflake for sure on the list, is evolve their platforms where it's easy to bring LLMs to the data as opposed to send Large data volumes to where the LLMs are.”

“I'm so sorry, but Salesforce is just a database. And LLMs are really good at talking with databases and querying data. You don't need 15 layers between you and that database to do it.”

“And to try to build a consumer tech brand and hold on to any population of users is going to be very, very, very hard in a world where innovation is almost free and LLMs can make anything.”

“Let's just say the money chasing all these LLMs, there'll be a lot of money vaporized for sure.”

“I think his point about like consumption pricing outpacing, or like, you know, ultimately taking the place of seed pricing, I think is going to be true. I think a lot of work is going to get handed over to LLMs over the next Five years.”

“And we actually torture test all of the LLMs. It's not yet the case that they're all equal.”

“I certainly think we will. And this may not be a popular position. Obviously, opening eyes, charging ahead, sort of leading the way right now. I think the market forces at work I mean, there's just immense energy to have an open source equivalent. Meta appears…”

“ColdCard was dead from like, 20, 12, 2013 to 20, 22, and with LLMs now, and tools like Orum, ColdCard is back big time.”

“A model trained that way is only as good as the smartest data in the training set. Like, it cannot discover new knowledge, because its job, the way the models are trained, is to do what a human would do in that situation.”

“The underlying thing that everyone's realizing now is that LLMs themselves are not a product. Like, an actual product is this entire software system that uses LLMs in it.”

“Frankly, I don't think startups today should be trying to compete on building bigger and bigger LLMs. It doesn't make sense. It's mainly a capital game.”

“There's still a huge amount of competition there, and I think exactly the same will happen with LLMs. They'll be innovating, you know, Google will add a feature, OpenAI will add a feature, some other startup that we don't even know about yet will add a feature…”

“We have big operational teams. We already use LLMs to augment those teams. Some places LLMs automate a hundred percent some of those tasks.”

“You can kind of see how it gets to the point that, that, that human, that person will also not be needed in a while as well.”

“I'm very sure that the outcome will be more favorable to us. Like if you think about how LLMs work, you need something to train the machine on. And the fact that we are, we had this discussion in paper, we wrote our thoughts down and they made an outcome. It's…”

“I have to say, I really like Jan LeCun's perspectives on various things, including his view that LLMs Our quote unquote off-ramp to super intelligence that, you know, in other words, we need a lot more scientific breakthroughs, as well as tamping down the fear…”

“Oftentimes I know people say, oh, well, AI is, it's just predicting words. That's all it's doing. Therefore it can't be intelligent. It can't be. And I think that's just. Demonstrably wrong. What I think is amazing, though, is the scientific fact that when you…”

“A lot of the underlying things that you need to do to be really successful in SEO. Are the same things that you need to do to be really successful with these LLMs, right? You need to have a product that is relevant. You need to have content that people find in…”

“And, like, it's not gonna be a 20 dollar a month subscription, that's not, like, the right way and it's not gonna be just, like, bundled API call, like, it's, there's gonna be much, much more sophistication and interesting models in that.”

“The real challenge with LLMs right now is that we lose all control. We make a request, and we get something back, and we hope that it is correct, and we're reliant on a very, kind of, unreliable review of what was produced to say that it's correct, and that's …”

“I think the rest are probably a couple years away, and as I said, I don't think the LLMs are the limiting factor. I think it's more the processes and the systems around them to make the LLMs more reliable.”

“I think the, we're gonna hit the maturation of the LLM in twenty-twenty-five, so I'll say that. I believe that LLMs are becoming commodities and that we're cresting the LLMs, that they're getting incrementally more successful, but not exponentially more succes…”

“I think it's gonna take next five, 10 years to, with the current state-of-the-art LLMs, it's gonna take next five to 10 years to get to very wide scale deployment of agentic plus automation.”

“Not because we are scared, because they are like idiot savants. Sometimes they can be extremely smart, sometimes they can be extremely dumb. And you have no idea right now how to distinguish between these two scenarios.”

“Every industry will be subject for a big change, but I also believe that we need a new giant leap in order to get there. I think, I don't believe that actual LLMs reason in the sense that I expect a guy with a 120 IQ reason. It's still, it's a stochastic engin…”

“So over the last two years, we we've spent a lot of time really trying to fine tune LLMs, build around them, and certain degree of success.”

“LLMs are actually not good at following repetitive steps. You are not gonna have LLMs multiply to numbers. No, you are gonna follow an algorithm, and you are gonna use script language, or you will program it, right?”

“I do think we are seeing the commoditization of the text generation layer, right? For most of the use cases that We'll see LLMs transform the world as we know it today. I think the current generation technologies are good enough. It's about, of course, improvi…”

“Linux has switching costs, and I think what we've discovered is LLMs have no switching costs whatsoever.”

“So when I look at perplexity as being perfectly positioned for the moment that the hallucination or really confabulation rate comes down because the moment that these models get good enough where you don't have to check the citations anymore, that's going to o…”

“Agree. But I don't think we are going to stay with LLMs forever. I think we're going to expand the systems that use LLMs.”

“I decided two years ago that I was just gonna read the attention paper and become an expert in LLMs, and, you know, just read all the papers after that, and I'm still in the process of getting anywhere near expertly status.”

“The reason that the writing from LLMs is terrible is because it's predictable.”

“Now, Google, ah, has to redo this because LLMs are better than search, right?”

“One of the reasons that these LLMs struggle to multiply large numbers is because multiplying is not linear. These LLMs could do anything linear without, you know, needing to think, but just like on a piece of paper how you need to write out all those intermedi…”

“As in LLMs are at that end, what we need is something that understands the world fundamentally, and this is the, it's JEPA thesis, it's called. I'm very bullish on this, but it's very frontier.”

“In the case of LLMs, for instance, you have these, what's called non transformer models that changes fundamentally the compute requirements.”

“Anthropic definitely has really stood out in, in their ability to solve programming and coding style tasks in a way that I think is like I'm very bullish on Anthropic for that reason, because I think that's such a groundbreaking part of what makes LLMs multipl…”

“I don't think any venture investors are going to make money from LLMs.”

“And so LMs are already good enough. They just need to be brought into companies to be actually made useful.”

“We monitor what percentage of search results are coming via traditional search engines and LLMs, and that's increasing. Still very miniscule, but the growth rate is, is, you know, shifting more and more towards these LLMs, and the way you have to optimize your…”

“Nine to 12 months from now, when the percentage of searches shift from Traditional SEO engines to these LLMs. A lot of the conversation for the next two or three years will be, you know, what are the best practices? There'll be tons of agencies popping up to f…”

“I'd certainly be a little nervous about how the code gen stuff is going to play out today, because it seems like that is just far and away the killer use case for LLMs and therefore the one that these big model providers are going to lean into next from an app…”

“But I think the undeniable fact is we ain't a monopoly anymore. It's not two or three companies monopolistically producing LLMs. There are four or five companies the other side of the pond cranking them out.”

“When LLMs finally work at something, the implementation will be boring as f***.”

“We've seen a five X increase, but it's still less than one percent of our traffic.”

“In 20, 26, you'll be able to open up a computer You know, log into North or whatever application you're using and say, file my expenses. And then the model will, yeah, figure out, you know, what expense policy it is and what the, where the photos are, like, do…”

“In voice, you still will see Pretty, pretty, pretty quick curve. On the lamps, I agree, it's probably a little bit flattening to some extent.”

“The way that LLMs work, they're very consensus driven. So they aggregate like everything that they've heard and create something. It doesn't create great copy, for example.”

“I think over time, we're gonna realize that LLMs are the telescope of the mind. That right now, they're making us feel really, really small. But in a hundred years, we're gonna realize that intelligence is more vast than we could have ever imagined, and we're …”

“SEO foundations are unchanged. And what I mean by that is some of the same systems that we rely on to get our websites indexed, our content seen, visible, crawled by Googlebot, and the rest of the internet are still the same systems that LLMs and generative en…”

“So in some domains, if you think like images, languages, like LLMs talking to each other at some point, you definitely get the degradation and that degradation is due to essentially like a loss of diversity of your data.”

“In terms of agents, you know, we worry a lot about hallucinations in LLMs. The parallel in agents is impersonation.”

“For us, base 44 was built in the age of LLMs, and so I structured the repository in a way that's going to be very easy for LLMs to navigate through and to read the code. I think for us, it's already closer to 90%.”

“Me personally, I don't believe we'll hit a speed bump. I think even with existing models, I don't think also we'll hit like a wall with like the improvements of LLMs, but I think even with existing models, we're only scratching the surface of economic value th…”

“Before LLMs, and I had an investment in a company that was pre-LLM customer support. Our resolution rate was around 23%. In other words, we could solve 23% of calls, maybe up to 30. With LLMs, with these new modern things, you can solve 60%, which means on a n…”

“If all the people are showing up In the LLMs, then all the advertisers are going to want to show up in the LLMs. And if you can help them get there, you can probably clip some of that money.”

“Now I expect reps to do their own demos. Why? Cause you can go into Grok. And of course I'm going to go with Grok, but you can go into one of the LLMs and we can enable these salespeople from a technical perspective in a way that wasn't really possible before.…”

“There's no excuse. We all use the same LLMs. There is no excuse for you to not have an agent as good as the new kids.”

“I think we are way under discussing the power of discovery in LLMs. I think this is the way we will buy everything in the future as we are embedded in LLMs. Well, I don't know why I would use anything else other than the best of ChatGBT or Claude or Gemini to …”

“It's all of the cute demos that you sometimes see by LLMs, like you can book your trip, whatever, or you can, it's really nice cute demos. Trust me. It's not even remotely close to being there.”

“Legal was a miserable category in SAS land. It's been amazing in AI land because LLMs manipulate language and lawyers manipulate language too.”

“I think what's going to happen with the autonomous agents is ultimately they're going to have an orchestration layer where they can have multiple different LLMs, you know, that they can orchestrate on and triage workflows.”

“I think if anything, we're going to see an excess of software being built on top of the LLMs, not less, exponentially more because the opportunity is so massive and the time is so huge.”

“I would have never had the confidence to push on the doctor so hard to get the scan if I didn't have LLMs open during the conversation with the doctor. I definitely would not have found another machine, got my dad scanned, and then figured out.”

“A prerequisite is we have like this entrance exam. There's this gigantic code repo that you need to read very fast. You can't read it with your eyes. You have to use AI. You have to use LLMs to ship all the features, and then you get a test at the end, which i…”

“Every 18 months, you know, the raw performance of the chip gets, you know, roughly three X faster. Then on top of that, other optimizations, how they run LLMs, how they do all the quantization, all the other clever stuff that, you know, you can read about and …”

“The thing LLMs have been somewhat unproven as of yet is the ability to actually trade stocks. You basically be a pod manager in a hedge fund world and outperform humans. The record on that isn't there yet.”

“Yeah, probably. I mean, I don't know about Anthropic, but on LLMs.”

“I think the analytic world is getting reshaped already where you can see people like Snowflake or Glean or Databricks. All these people boast enterprise data lakes where you can bring the data and run LLMs against it and get you much more synthesized analytics…”

“It's a moat destroyer when LLMs will lift you from one vendor to the other, and I, you know, I don't even, like, I literally was doing a pitch this week, and the founder was going on and on about their moats, and I immediately didn't want to invest. Like, I ju…”

“The main thing here is that LLMs are really amazing at code, and the amount of economic value that can be created from that dwells to everything else, right?”

“In today's world, the words that you say and how you portray something, that becomes a part of the story that the intelligence systems that we're building ingest. And they, they're like world model, LLMs, and the tools that we're going to use to do work on a d…”

“One way to think about LLMs, right, is they're just, they turn energy into, like, GDP, right, or into revenue.”

“As opposed to having humans in your compliance and security team over time, label data and decide what goes where and what can be accessed. I think we'll just start to trust LLMs to do that.”

“I think the LLMs will make many fewer of these mistakes than humans pretty quickly.”

“Anything that can be reduced to code will be done by it.”

“The only thing that mattered for the last two years is LLMs do code, and code is a half a trillion dollar industry.”