Straight Answers

“So how do you, how do you make people happy?”

answered: “the way to think about agents and the things that the models in vZero need”

“Kind of like math problems. How does their approach differ from what it is that you guys are doing?”

answered: “I think the sort of concept formal theorem proving actually, uh, existed”

“Yeah. And I'm sure we're going to talk about it, but obviously the question is, uh, to which extent can AI review AI?”

answered: “all of the problems that I'm talking about are very much our opportunities as investors”

“Uh, but what, what does that mean in practice? If you're, uh, a developer at Vercel, and, you know, you want to do great at your job, and Guillermo says, well, you need to have better taste, what, what do I do?”

answered: “Maybe just to set the background, I believe that every big intern internet or technology wave”

“So, starting with models, uh, so you sort of offer power hundreds of models. What is the latest number of Models that you, um, that you offer?”

answered: “Yeah, so our product has multiple layers. Uh, at the lowest layer”

“That search problem, you solve it on a per-task basis, so sort of a priori before the, for the specific use case, or are you going toward, or maybe you do it today, towards real-time optimization?”

answered: “when we think about, Uh, this optimization, we also have two modes.”

“So those are the suggestions when you route the right answer to the right person, those are just suggestions, and then the, the physicians makes the call?”

answered: “Think about the world as our QA system, as a QA team”

“Um, how has the transition to being a CEO been? What was surprising, different, challenging, or perhaps not so challenging?”

answered: “Well, I wouldn't say I have this illustrious background. Let's start there.”

“broad group of professionals, you know, smart people that are good at their jobs, trying to be good at their jobs, and some of them will be doing revenue ops, some of them will be doing marketing, some of them will be lawyers, some of them will be accountants. What does taste mean in a context where you have such a broad audience, and how do you test for it?”

answered: “I think a lot about the phone and how all of us start with the same phone”

“Yeah, how does, how does it mean if you say no to, like, features or?”

answered: “That's actually one of the biggest challenges that we have, especially in the product team”

“How do you balance useful research with great exploratory research?”

answered: “My belief is that research needs to be bootstrapped.”

“How far do you think we are from that world where we have armies of it?”

answered: “we are at, like, you know, the very, very beginning tip”

“I have access to co-work in my personal life, uh, but, uh, not, not at work. Is there a, uh, roadmap for that?”

answered: “I can't particularly, like, comment on anything that we're currently working on”

“It feels like a great Uh, place to start. What's your assessment, uh, about where we are and how wide the gap between closed source and open source currently is?”

answered: “it's really exciting to see all of the energy going into open technologies”

“Going back to the jump in, uh, performance from last year's models, or even this year's models, or even actually Sonnet to, like, 4.5, uh, again, to the, to the, uh, point about the, the pace of progress accelerating. What, what were some of the, some of the breakthroughs?”

answered: “That I can't really talk about. Yeah, I, I mean, I think,”

“So in, uh, AI video, there's, uh, this well-publicized, uh, you know, debate around, uh, in particular using YouTube videos, and there's, uh, class action lawsuits, and all the things. Where do you all stand on that, and, uh, what data do you use to train the current version of the models?”

answered: “we don't disclose what data we use, but we've done some”

“Uh, what does that mean in terms of of capabilities? Could you perhaps give us a little bit of a tour of what the, uh, whole, uh, aircraft carrier looks like?”

answered: “goes to your other point about how have we evolved as a company”

“Switching to, um, a deeper dive into AI, which we've covered a little bit already, um, what's your sense of the reality of the markets in from the perspective of Snowflake customer, um, you know, in particular, the question of, uh, people going from, from POCs to actual implementations, where, where are we?”

answered: “I think it's useful to step back a little bit and talk about the priorities”

“So scaling laws, um, to continue with very easy questions, which have a clear yes or no answer. Um, AGI, ASI, what's your sense of, um, how far we are?”

answered: “now I'm like at 25 pages and like, oh my God, it's really, really hard to define.”

“How do you guys think about it? How autonomous are those agents, uh, in, in your deployments?”

answered: “To me, the better question usually is how much you trust the agents”

“Do you think that's a world where, where, uh, that's a temporary issue, you know, a little bit related to the earlier part of the conversation about, uh, AGI, SAI, or whatever, um, is there a world where, uh, I don't know, the underlying models improve so much that, uh, this ends up actually working?”

answered: “I don't really make predictions, you know.”

“So, uh, you know, starting with 24, but I guess the structure is the same today. So there's, there's, uh, there's a private version, there's a public version, there is a paper prize, like, how does that all work?”

answered: “Some of the goals that we have for dark prize are to one initially”

“Why did you switch? What are the pros and cons?”

answered: “where do you think this is all going? Like, what do you do with”

“What a, what a, what a concept. You meaning you, you don't burn five million a year?”

answered: “it's quite unusual for AI startups”

“Great. So, um, back to, um, Victara as a business, so, um, how do you sell, who do you sell to, uh, how does the, yeah, use cases, how does the business side of the story work?”

answered: “Will was talking about open source, I have went through open source”

“How do I do that? How do I take, like, video and translate that into numbers?”

answered: “we can talk about the mechanics of how to do it, which is maybe the less interesting thing”

“So what, what does that look like long term, something like Crypto, CryptoKitties? Go into crypto dogs.”

answered: “I won't speak specifically about them, but I'll speak about why I think the area is interesting.”

“It's like this concept of like digital transformation, which is amazing to me that, isn't that what all of us have been doing for the last 30 years?”

answered: “it's a journey that I don't think ever totally ends.”

“And, and, and I read somewhere that you guys had a project, and I don't know if it became a product, but that was, ah, Talk to Doc, was it like almost like a Siri for, ah, for, to, to, to use an LP, basically, to, to query the database? Is that by voice?”

answered: “the mission that we have now, now that we have all this data”

“This Spark, this data flow, and, you know, maybe MapReduce is not what people need, and all of this is evolving. Where do you think this is, um, evolving from a product standpoint?”

answered: “let me talk a little bit about sort of the, the, the market status right now”

“And so, uh, when you compete with the other Hadoop distribution vendors, what's the, um, what is the positioning? Is that a more mature ecosystem, a better editorial choice of what technologies matter?”

answered: “I don't want to talk about positioning versus competitors”

“How do you, I guess, how does that work? And how do you think about, uh, what needs to be in the open source and what needs to be in the commercial product?”

answered: “my answer is going to be a non-answer. Um, uh, I, uh, we are still early”

“And it seems like over the last few months, it's become like a, like a, Uh, huge area of, uh, activity and, uh, with a lot of people competing. So what happened to that space?”

answered: “just to bring us back to Atlas for a second”

“And, um, I think I heard you somewhere, uh, because I think that was part of one of the several jobs you had, um, at, at Activision, that there was a security component as well, like a, uh, intersection of, um, AI and security. What is that, uh, application or use case?”

answered: “let's speak more generally so that I don't get in trouble for my trade secrets”

“This has been shown to strongly sway preference in the context of political candidate selection, for example. Does Facebook ban the personalization of advertisement along this axis?”

answered: “I don't think I can answer, uh, That's a really tricky question.”

“That seems to not be the case at all, uh, in 2026. Uh, can you walk us through some, some ideas for what is happening in pre-training and why it's progressing now in a way that people hadn't predicted, uh, last year?”

answered: “I can't talk, um, in a lot of details about what is happening internally.”

“community that's wondering whether open source as an ecosystem, not, not Nvidia, but in general has been progressing in part based on the ability to distill closed source models and in a world where we seeing the anthropics and fable Fives of the world starting to discourage distillation. Do you think there is a chance that open source AI progress may slow down in that context or as a result?”

answered: “community-oriented approaches to developing and deploying AI are gonna continue to strengthen”

“37. How do you, um, train a model to encourage the model to do that kind of things and come up with brand new ways while being efficient and exploit known path?”

answered: “I actually have a, I have a funny story about this.”

“Extremely important. You mentioned, uh, the various teams, uh, at OpenAI around, um, safety, security, uh, can you, can you provide a bit more color about, like, how that's organized internally?”

answered: “main point I want to highlight is not the precise sort of structure of those teams”

“So I think you, you guys started as a, um, uh, transaction, there was a transaction layer was the first one where you would, Connect with a bunch of bank accounts, extract data, and then enrich it with, uh, categorization, localization, the merchant, the category, you know, the categories, that type of thing. So that's, that's still the core, and then what did you add?”

answered: “Actually, the name, the name Plaid came from an early algorithm.”

“Isn't a GPU a thing for gaming as well?”

answered: “Oh, right. Yeah. There's, there's also that, right?”

“In 5.1, you have additional granularity in terms of telling the model how long it should think. By default, how does the model decide how long it should think?”

answered: “the model sees the task, it will decide on its own a little bit”

“What creates the conjecture? Is that front-based, um, like, what goes into it?”

answered: “the conjecturing part is still the underdevelopment part. Like, we have been... focused on Prover”

“Let's start from, uh, the top end for anyone that, uh, is just starting to learn about cloud code. How do you describe it?”

answered: “5070 years ago, the way that people programmed looked very different”

“Uh, so how do you all think about AI as an opportunity, uh, or possibly a threat? Have there been any moments where, you know, uh, despite being bullish on, on AI, um, you know, You all look at each other and say, oh, this one actually might be a problem.”

answered: “alert to the rise in visual AI, but we feel we're actually really well positioned.”

“What was that story? When did you come in, come in, and what was the context?”

answered: “the idea that drives the company today is the first idea”

“Is that more of a media creation? How do you, what's your take on it?”

answered: “what we do is we contextualize the language models so that they can do their job”

“Is it, uh, like a completely different set of tools to be able to do that? Or, or, or, or does some of it come from the, you know, the, the, the very nature of the, um, of the media property?”

answered: “So you're talking about like, what can be, um, what, what makes it hard?”

“And without naming names, can you, can you talk about, um, any kind of examples of what people have actually, have actually done? It seems that, uh, so far we are in the kind of like low hanging fruit, kind of like easy win part of the market, um, cycle where people do things like search and chatbots and that kind of stuff.”

answered: “what's interesting to me are the things in the middle”

“What is it? What does it do?”

answered: “These are great use cases for us that produce a ton of value.”

“Do they, uh, take some of the inbound and just like have that first conversation? What, what, uh, and I, I know you're figuring it out, but like, what do they do currently?”

answered: “Yeah, I mean, there's a fair bit of that.”

“And one, is it fair? And, and, and two, if that's, um, what you're saying, then Uh, for AGI using the current paradigm, basically what we just described in the last hour of pre-training plus RL gets us there.”

answered: “I think the AGI word is actually pretty not useful.”

“So let's jump right into the overall vision for the product. What is it that you guys are building and why did you choose that specific problem to go after?”

answered: “I think it's worth taking a step back and kind of where Jason and I found ourselves”

“I mean, it could, could it be the, the model provider underneath because like the agent went haywire? I guess all of this is going to take time to work this way through the ecosystem.”

answered: “When it comes to sort of payments in particular, you, you mentioned a bit ago”

“How do you build the taste that you need to be able to review those things? It's a complicated topic.”

answered: “The idea of prompting, too, is just fascinating.”

“You know, it sort of feels like last year AI was about to kill us all, and it sounds like this year we're all like, okay, well, maybe my POC can get budget approval kind of thing. What's the reality of it?”

answered: “I think what's happened is first off, obviously chat GPT, Went mainstream.”

“One will accept working with you and two that, that partnership will make a difference for you. Were you at the right moment?”

answered: “partnerships, Often don't work. Like, 90% of them don't work”

“And, uh, so now with Deep Blue and reinforcement learning, then, um, how does that, how does that work?”

answered: “So the challenge for a lot of organizations, and this is sort of how DifBlue got”

“So, you know, you mentioned graphics, uh, and then you mentioned the specific use case of security, but like, what, what, what do people in, uh, at a, at a place like, uh, Call of Duty slash Activision do with General AI? Do they experiment with it, play with it, uh, whatever you can talk about?”

answered: “I'd say more broadly at, like, what I've seen publicly discussed at other gaming companies”

“You mentioned, you know, the temptation to verticalize or, or, or, or, you know, or not like, how do you, I guess, how do you think about it? How do you sell, uh, what's your go to market motion?”

answered: “biggest driver of new users has always been Mystery. We call it like direct.”