2035 Decoded: Navigating the Decade Ahead | Dust, Glean & More | RAISE Summit 2026

0:04 Hi, good afternoon everybody. welcome to the last session of the summit. I've got a fantastic panel here joining me. so thank you very much for being here. So I I would say I've been to a couple of panels this week where one of the questions has been where are we going to be in five years time or where are we going to be in in one year's time and the answer has always been that's completely impossible. I have no idea. I don't know where we're going to be by the end of the year. So, we're trying to tackle the 10-year view out to to 2035.

0:36 and one of the reasons I think we need to do that is because the other thing I've heard this week is that enterprise adoption of AI is a 10 to 15 year journey. This is not something we're going to do quickly. and a lot of the leaders who are the customers of the companies here or here listening, you know, they've got to think about that time horizon. and they've got to make decisions now that that have that that time horizon ahead of them. And so what we're going to try to do today is to not try and predict what's going to happen in 2035, but we're going to try to think about some of the assumptions, some of the ideas that people will need to consider as they go go through that decade ahead. So we're going to ask a few questions about technology, about people and about organizations to hopefully challenge some of the challenge some assumptions and give people something to think about. as we go forward and on that basis if we start we're going to start basically with where we are today and I'm going to go to Eve at the end there first to really say you know what when we think about assumptions that drive the future what are we getting right what are we getting wrong how are people thinking about how should people be thinking about where we are today as a basis for going forwards >> can you hear me yeah thanks for having me that's a great question. so to give you an answer, I just want to define what is AI because people say AI AI but like what is actually AI and the answer is interesting because like 10 years ago we were dealing with neural networks and you know data sets and new knowledge were defining how good and neural networks are. six years ago we started talking to to frontier models companies and you know frontier model was the form of AI. three years ago people start moving to harnesses. So it's something which prompts your foundational models and make it smarter.

2:29 So we call it AI too. And to me AI today is all of it. So you need to have foundational models, you need to have harnesses which prompt your foundational models and make them smarter. You also need to generate new knowledge which you feed back to your models. So it's like a spiral or a loop which feeds into each other. And when we speak about foundational models, people always think about LLMs.

2:59 Yes, historically LLMs are the first form of AI large language models, but reality is like there's so many different architectures nobody talks about. So with logical intelligence, we're working on energy based reasoning models or sometimes we call it latent reasoning models. those form of architectures designed specifically for spatial reasoning for prediction and planning. So like Amilab Zanun he's speaking at the same time with us I heard he's on our board and they also building EBMs just different kinds for robotics and we build an EBM specifically for financial analysis data analysis for mission critical industry where you cannot have a mistake so we speak a lot from the companies in Wall Street and Wall Street don't want to share their data with charge or other frontier models they want to have their own form of AI which is fast trainable specifically designed for the problems of their business and something which doesn't make mistakes. So LLMs are great for language they're great for poetry and you know just like therapy discussions for people but it's really bad when it comes to formal reasoning so it's just not designed for formal reasoning. So the EBMS are specifically the models for formal reasoning. they're not good at language but you know sometimes we don't need the models who who is going to be doing everything. so I see that back to your question we need different ecosystem of different AI models. I also see eyes as like people you know some people more inclined towards formal reasoning some people more inclined towards towards arts and humanities and variety of people what makes the society better. So we need a variety of models for different purposes.

4:52 >> Thank you. well we I think we'll go around if everyone just to to say a quick intro and and what are we what should we be confident in and what is the world getting wrong right now? Maybe go to Raphaela next. Sure, it worked. Perfect. hi everyone. Great to be here. I'm Rafael. I'm the founder of Discontinuity Capital. It's a market neutral hedge fund. I invest around the agentic AI dispersion around a framework that I will explain probably later.

5:21 I think what people are getting wrong today, including this morning is that we are at 1% of what is happening. Just last week, everyone was freaking out about AI capex and in the public markets, everyone was saying, "Oh, the hyperscalers are going to die down and everyone was stressed." And this morning, Meta announces that they're actually doubling the number of gigawatts of compute. And so Meta, which was spending $75 billion in capex in 2025, is now estimated to spend anywhere between 250 to over 300 billion in 2027.

6:01 So back to where we are, we are at the very beginning of the AI buildout. Aentic AI, we will talk about during this panel, is a very sophisticated architectural construction. And it's actually hard to have a genetic AI systems work because of the models, because of the hardness, because of everything that involves this new form of living system. And we have a long way to go in my view. >> Alvin to you.

6:32 >> so we're introducing ourselves in the company. Is that right? Or just answering this question, what are we getting right or wrong? Sorry, I I I >> just a a quick hello and then your your topline view on on what are we getting right and wrong. >> Okay. Okay. Got it. Yeah. Well, so my name is Arvin. I'm the founder and CEO of Glean. Glean is an enterprise AI company. think of it as a more stronger version of Chat GPD inside your inside your enterprise. a system that's connected to all of your enterprise data and knowledge all of your enterprise systems and also has access to all the great models whether it's open source or or or these clo closed form foundation models and it uses the combination of those two things to then help your employees ask qu you know answer qu you know get answer their questions or do some work for them so we are a enterprise co-worker and assistant and we're also a platform an enterprise AI platform that that actually like again like you know by being connected to all of your enterprise systems brings the right context the context graph to all of your other AI applications and agents that you're building in the enterprise. So think of us as a AI platform and an AI coworker. Now in terms of like you know top you know views on AI what's what's what's right what's wrong I I I think the there's a I I would say that you know there is immense focus on on ROI with with AI and and sometimes you know it comes prematurely like I think for any enterprise the right now it's clear like you know as as a human you would understand today that this technology is powerful and you have to make it part of your work life and so I think we just have to continue that journey without being you know overly focused on ROI but just do it in a way where we're being smart and efficient about it where we use the right models for the right task where we don't sort of you know burn through and do things like token maxing so that's sort of like you know my view of like where we need you know where we need to be headed you know with AI.

8:35 >> Thank you Gabriel. >> Thank you. Hello I'm Gabriel the founder of Dust. We're a series B company that serves fastm moving teams with a horizontal platform to design, deploy, and maintain agents and really empower human agent collaboration in a multiplayer way. I think a lot of people who have chat GPT or some solution at work realize that they're very much on a single player journey with people downloading skills from X and feeling proud of themselves. But we believe that the real next level for enterprises is to actually have a system that allows many humans and many agents to work over long time horizon tasks in a very collaborative way. So customers like Vanta, Data Dog or Decagon for us are leading the the charge in that way. I think what we're getting right is that the transformation is big. I think what we're getting wrong is that the transformation will really very rapidly increase differences between winners and losers. And it's hard to over to underestimate, but I still think we're underestimating the the delta that is going to occur. In the same way that self-improving systems for models are likely to give some of the leading labs an advance that will be hard to catch up with, I think that the companies that have set themselves up to organizationally compound are going to take off at a pace that is hard to overestimate. So what we're seeing in the way the smartest companies are starting to really leverage in a deeply both secure but also collaborative and and and across functions way the pace is just changing a little faster than everybody expected and I completely agree with your comment on the fact that ROI is is a relatively trivial one to to solve once you've understood what is on the other side of that transformation.

10:29 Yeah. >> Hi everyone. >> Close. >> Okay. >> Hi, I'm Nico. I'm the CEO of Corgi. We do AI for insurance infrastructure, financial infrastructure broadly. We're a full stack AI company, servicing thousands of customers. I I think it's very easy to to get lost in in kind of the weeds of of talking deeply about the specific technological innovations or the financial kind of implications of what's going on with with AI. you know like it's is very important to understand the capex or or model capabilities. But I I think one thing that people tend to get wrong u pretty frequently is very much either over or underestimating the rate at which white collar work is actually substantially changing with AI and what those changes might actually in a practical sense look like.

11:28 is pretty clear I think to most of us in this room that that programming for example writing code by hand has very substantially changed and and permanently changed in the past you know 6 months which which probably has very long-term implications on what what that sort of profession looks like but kind of applying that to to other industry segments is something that is often times not really appreciated and that's something that we see every day I mean we run an insurance carrier and reinsurance company with with without human underwriters in production with a minimal amount of claims adjusters with these sort of roles automated away and and throughout the economy you know there will be pretty big changes and I think on the short term people tend to really overestimate the capabilities of of what frontier models can do. but on the long term are probably underestimating quite a bit how substantially different the the very big sectors of the economy that that comprise a lot of white collar work might end up looking like in in 10 years.

12:31 >> Brilliant. Thank you. And there's a lot there that I think we can can dive into. So if we go from sort of diagnosis to decision- making we're going to start with technology and we're going to go to you first Arvin. You know, if you're a leader making technology choices now and you're trying to think on a 10-year horizon, maybe you're an infrastructure company, you're you're a bank, you're you know, you've got that view where everything's changing so fast, what can companies build around confidently.

13:00 and what should they be wary of or avoid maybe committing themselves too quickly? so this is the this is the area of technology where change is so rapid that everybody whether you are enterprise you know company buying technology products or whether you are an AI company building technology the rules have changed the game has changed and you have to sort of figure out how to actually operate in this really really fast changing landscape for enterprise buyers, some of the most fundamental decisions that they need to make today are how do they how do they remain in control of their intellectual property, you know, and in this new modern AI world, how to not seed control to to the tech vendors on fundamentally how your business actually works. with AI in the future, a lot of the work is going to be done by AI agents and those agents get better over time as they get feedback from humans.

14:09 and they're sort of building their own memory of like how to actually do any given task with the highest quality. So now the question for enterprise buyers is that as those learnings compound as AI learns how to do work in a really really nice way who has all of those learnings who owns all those compounding learnings and the answer should be clear that it it's enterprises need to own it themselves. So from that perspective there are a few decisions to be made today. Number one do not commit to any particular model. You don't know which is going to be the right model for you know for any given task. In fact, there's a lot of divergence in AI models. They're all going to be, you know, better at different things. So, make sure that your strategy is such that you're able to enjoy all the model innovation coming from all the different AI companies as well as open source.

15:02 second thing that I would say from a architectural decisioning is instead of having like an fully decentralized strategy for AI where every department every function gets to decide what their AI stack is and what agent platforms do they use. you have to sort of figure out the right reference architecture for your enterprise and what layers needs needs to be central and managed by the code tech team and what layers can actually be distributed to your different departments. So so from that perspective one of the key technologies today that all enterprises are looking to invest in is context graphs and so so how do you build that? That is that is one area where enterprises should be making a choice for who are the right vendors and you do have to pick one and and and sort of go with them but still again build it in a way so so that you know you're able to replace that vendor with you know somebody else. So ensuring that you know they all provide standard interfaces to interact with those context craft is going to be important. And then finally I would say the most important thing with you know coming back to those compounding learnings with you know with AI every enterprise is going to need a memory infrastructure.

16:27 Memory is at a personal team departmental agent level and you have to you have to make sure that you make investments in a way that allow you to retain control of that memory within your enterprise. In fact, it should not be leaked because this is going to be your key sort of differentiator and competitive advantage in the future. So that's sort of the principles you know with which you have to work with in in my opinion and and that sort of you know helps you now make you know the decisions on like what vendors to choose for each one of those layers. it's not so much so important to actually pick right vendor right now. We just but what is what's more important is to have an architecture where you can switch where you can actually switch from vendor one to vendor two and you feel pretty good about like doing that based on the the the sort of the architectural choices that you have made.

17:17 >> Brilliant. And and Gabriel you talked a little bit about the big gaps bigger gaps between winners and losers. How do how can people think about navigating a world where some people win big and some people disappear quite fast? I I believe one of the angles that companies should obsess about is is really the density of the talent that they have on board to navigate the years ahead and the way in which they have made decisions especially for large enterprises over the last decade or even the last years has radically changed.

17:50 They need to first of all realize it and then take the consequences for themselves and their teams. The people who will do incredibly well in the decade to come are generally more curious, more confident, think in systems and more transversely and are able to extract value from these systems much more dynamically. In a company, you're either building or selling or doing something that helps one of those two jobs happen.

18:20 If you're selling, you're in a job that I would consider a run job. you're trying to get these new prospects and new customers as much as possible in a zero sum game against other companies trying to sell. You need to be working at a company that has the best possible equipment for you to do that so that you have the highest chances of selling well. And I think that even those run jobs are going to increasingly see value to having some build where the best sellers have an idea of how they can make their job a little better day in day out. the companies that hire those bestsellers that have that systems thinking and give them tools to propagate that tip, that trick, that new technique to their team the day after it was invented are going to compound. And in the build jobs, software development or roles adjacent to that, it's quite obvious that you have to increase talent density because the amount of leverage these humans have in the years to come is orders of magnitude higher than the amount of blast radius that they had a couple of years ago. So a mediocre engineer with 10,000 agents is going to wreak havoc in your codebase in a way that's very different than a mediocre engineer would be able to wreak havoc in your company in the past year. So I I just think about the pull requests that agents are making whether they're reviewed by humans or not and the amount of noise that is going on when the architects and the builders behind that do not have the foresight, the perspective and just the sheer quality of reasoning. So that's one of the ways in which I think that change is happening very quickly and I'm still surprised to see that some companies haven't revamped their interview processes more drastically to take into account how capable are you to adapt to this technology. You were making a comment eve about you know making mistakes or not. We've moved to a non-deterministic set of tools. Now traders have been making money with non-deterministic tools for a long time.

20:10 I started my career as a quant. It's fine to be okay with risk. But very few people were trained to do that. Many people learned their jobs as if the machines they touched were like typewriters or calculators. You punch the same keys, you'll get the same result. The fact is some error rate in everything you do is probably fine if it can be in exchange for an amount of upside that actually is efficiency or speed all those things and moving your task force to a non-deterministic mindset is that challenge that's what I would encourage >> I think Eve's desperate to respond to that >> okay I want to comment on deterministic side so I understand that everyone is excited about scaling like putting 10,000 agents to do you know very fancy work but nobody really talks about the safety like how do we know this agent is going to do what it's meant to be doing especially if you LLM LLM is a language based model which is based on words it's going to try to guess words what word is going to go next and the model in French is going to be thinking differently from the model in English so you have language dependent intelligence literally which will hallucinate just some words going to be naturally next to each other it's going to produce some hallucinations and imagine you put in LLM in a banking system which is going to move billions of dollars for government or for people and your language model is going to make decision oh how this smart contract is going to perform and the answer is you know we don't know it's going to perform or not like you know it's going to be hallucination and sometimes billions of dollars going to move to another bank account or you put an LLM in a car in a self-driving car and somebody says oh you know 20% is going to drive you to a forest just because this word next to that word, you know, can control it. So, I feel like in 10 years, we really have to think about safety and measures, how to constrain AI and forcing it to behave the way it's meant to be. So, one of another example was mentioned here was VIP coding, right? So, everyone is excited to generate millions of lines of code. but this is how our company gets customers. So, we we're speaking to banks and banks like, "Oh, I hired an intern. They in turn moved my legacy code from one language to another. They used some frontier model to conversion that there's no logical equivalence proven between this conversion. So it means it's buggy. And so they they bring it to senior departments and they say well I don't know this full of bugs. We we generated millions lines of code. We don't know how to debug it anymore. How about we have another agent to debug it.

22:42 But another agent is also another LLM agent which gonna bring hallucinations and make mistakes and so on. So at the end who is dealing with this bank and government regulators who's going to say hey can you give me a certificate that the you generated is correct and the answer is none. So they come to us and we provide formal verification. It's a technique which constrained the LLM output in a verifiable form and it's machine verifiable. So you don't have to debug it anymore. So if you vipe code, we're going to make sure whatever you vip code it is correct. And there needs to be more companies like us who making sure your LLM is going to behave the way it's meant to be. Your agents behave, it's meant to be, your code generated, it's meant to be. So in 10 years from now, this should be the number one thing because if you put AI everywhere, this AI is going to do crazy So >> Okay. Do you want to have a quick response there, Rafael?

23:38 >> Oh, sorry. No, it's okay. >> Okay. >> Humans are making mistakes every day and many companies still manage to make a lot of money. So, it's it's all about the nuance where security is critical and formal software in nuclear power plants or black boxes and airplanes has existed for a while for that reason. But I think the appreciation of where you can tolerate more risk is probably going to let some players move ahead very very fast.

24:02 >> Thanks. I I think like I take a step back and I put back my head of like okay let's look at this as an investor and think about what is at stake from a value perspective. So first we've heard that we are moving from humans being the main actors to agents and potentially swarm of agents like lots of them. Is it one to 100, one to a,000, one to 10,000? We don't know. That's a very big change for any enterprise. Then we're speaking about going from deterministic to non-deterministic and having this margin of error that we have to determine. Now imagine like how profound that changes on any company within the S&P 500, the Kaharon, the Footsie, whatever. Like imagine if we take the sector of insurance like this AI native insurance company, this AI native bank versus AXA, Chub, JP Morgan, etc. Can you imagine the difference that is at stake between one company that will actually be able to make this transition into the agentic era and have these attributes and the profit and loss statement that goes with that? Meaning the better margins, the control of inference costs because the companies will not be token maxing and will actually have the right architectural control, the new revenue lines that are possible. like we are talking about the biggest value dispersion that we have seen in decades.

25:29 I personally think much bigger than the internet. I don't want to go into comparisons but I just want to underline like how the technology connects to the economics and to the financial value at stake and this is absolutely massive and I think that while we are still figuring out as you are saying what are going to be the best models per which use case etc we don't know like things are evolving very fast we've made a ton of progress on agentic AI we think that oh now all models are equal but I can tell you that the frontier still matters and there is still a gap even if it has narrowed significantly between frontier models and open-source models and that gap is what allows this whole race to continue and so I think that and I'm seeing this in the public markets because all of this is happening in the private markets all of these great companies here are in the private markets but the public markets are still figuring whe whether this is real or not I think everyone agrees that this is very real, but now it's time to actually think about what is at stake and how we rethink the value of all our economies.

26:41 >> Okay, brilliant. Well, I'm going to slightly jump the we I've already blown all the time limits here, so I'm going to slightly jump the order. Gabrielle, you talked a little bit before about about the importance of talent density and people. I'm actually going to take that question straight on to you, Nikico, if I can around how pe how do people think about people over the next 10 years? Do you or are you designing organizations around people? Are you designing organizations around the two together? Or are you designing machine first? If you cast your mind forward, how do you how would you encourage people to think about that?

27:12 >> You know, people are still really valuable. It's just they're valuable for different things. you know, if you go back in time and you look at banks, the number one employed kind of profession, or employment category in banks used to be computers. And I don't mean the machine. I mean there was a job title called computer and they were people that wrote down you know how many deposits and with withdrawals there were in any given bank. It's a very important job but today banks employ approximately zero computers. A computer in fact in our minds refers to a machine and there's a lot of kind of categories of work where this sort of repetitive you know checking the box of various tasks is is is luckily going to be automated and that's a really great thing. And what we can see with large language models is that a lot of words based tasks very importantly are going to be automated. And over the past 50 years before this decade the big story was a lot of the number space jobs became increasingly automated. And what we saw there was was not the removal of people from from you know sectors that have numbers but rather a huge productivity gain where suddenly with with digital spreadsheets with with computers suddenly a lot more people could be supported in terms of you know working with numbers all day and in professions like investing like accounting white other white areas of white collar work and I think similarly there will be a renaissance within a lot of language-based industries simply because the efficiency per person will go way way way up. so at least at least for me I think that like every business becoming a lot more profitable, a lot more efficient, saving a lot more time will naturally lead to more hiring in in high leverage areas and the historical precedent is actually quite good for that with regards to to to previous errors of automation.

29:12 so, you know, I I don't worry too much about about cutting people out of the loop because I think if people can be automated away, then, then that means probably, their job wasn't that great to begin with. >> Okay. So, the human stays in the loop sort of. Gabriel, you do you have any response to that? >> No, I think I I think I agree with the the fact that we've overfitted some appreciation of roles for things that other systems can do. Nobody's sad that most of the mining is automated in most of the planet. nobody's sad that mo that most of the dangerous exploration jobs for petroleum exploration are now automated. Everybody's happy that machines are doing that. So there are obvious edge cases that I I think that it comes back to trying to understand how res-killing will work in very large companies and maybe the disadvantage at which large employers find themselves given the velocity of the change. so one of the ways in which we've designed our platform is to try and act as heat seeeking software for empowered talent by allowing companies to identify and I think it might happen also w with others on this stage identify who's reacting very well and surprisingly well to this kind of technology who's really at ease with it. It's very hard to measure because it wasn't on their performance scorecard historically and in many cases they're actually able to drive transformation at the t the team or department level in ways that are super surprising. So that's one of the angles that I think is is really very important. We're not going to be equal in the face of this transformation but companies have control over how much understanding they build right now about the quality and the resilience of their team to these changes.

30:53 And Arvin, you think a lot about organizations and the shape of organizations in the future and we've heard a lot this week about the need to transform whole organizations or whether organizations should should start from somewhere else and start somewhere new. How do you think about transformation across this time frame? >> Yeah. I so I think like first just like everybody else is saying people are the the are the key and you know like that's what the organization really is like I actually still fundamentally believe in not putting AI or AI agents at the same level as as people inside your company like you in in my opinion you know the the hierarchy is actually very clear AI is simply a tool in our in our tool belt and as humans, you know, we actually use these tools to work differently in the future. so from that perspective, I also don't think that there's any like massive acceleration of some jobs getting eliminated and others getting created. I think mostly I'd see it as a incremental journey where almost all job functions get a little bit more exciting that repetitive works work actually goes away. and and the roles may got get a little bit more broader like you may reduce like for example today it's easier easy for an engineer to also act like a designer or a product manager to also sometimes you know be be the engineer like you know when it comes to prototyping systems so you will see like people's ability to to do to work actually is going to increase because of AI but as such when I like when it comes back to designing an organization. Well, I mean, you know, you have to design it 100% around people. That is it. I mean, I think that's that's my opinion. yes, like some of the things may change. For example, maybe maybe because you have great tools available to you, you can manage more number of people compared to like how many you can do today like you know in your job function. So, that's that's the only change that I'm seeing personally like you know happening in in organizations.

33:05 they may get a little bit flatter yeah the the the the scope span of control may actually get wider but I think we will still continue to have hierarchical organizations as as opposed to like you know there's there's there's a lot being said you know in the opposite direction of it >> which takes us on so nicely to the end which is we're going to end up by looking at value and we're going to go to Raphael you talked earlier on about this is one of the greatest dislocations of value we've ever seen. I' I' I'd love to hear your forecasts on that, but I won't ask you to forecast on that. What but what are the big what are the sort of the second and third order effects that people can be thinking about about where that value is going to go and and how is the market going to price this transition as it develops?

33:55 >> Sure. Well, this is my this is my favorite topic obviously. so 500 companies in the S&P 391 companies are non like AI names in the sense that let's put away the semicon the semis companies all of the companies that have made everyone have a ton of money over the past year the Nvidias the AMDs all of the companies the companies in memory etc let's put those aside for a minute and let's focus on the companies that are the banks the hospitals education average software software we've talked about a lot about software last year that gives us 391 companies per the framework I have built I think that there are going to be out of the 391 companies approximately 250 companies that are going to lose sorry and 125 companies that are going to emerge as winners and the companies that are going to emerge as winners have one thing that is their moat that you were speaking about which is context. Context in my view is what allows a company to have superpowers in this agendic era. JP Morgan has context, FedEx has context, AXA has context, BNY has a ton of context. Context is data that is deep, that is defensible. You cannot go and build a FedEx FedEx across 50 countries overnight, even if you try very hard. You cannot build the custodian status of PNY overnight.

35:27 There's a lot of things that you cannot do. And when you make that data accessible by AI agents so that they can actually transcend your organi like translate your organization as it as it is currently into this new reality. All of a sudden you find yourselves with companies that have superpowers. This is really the term I like to use. and who can imagine being radically different than what they are doing today. And we go back to the ROI debate. We have to stop limiting AI to oh I have a better margin. I was able to cut my cost etc.

36:03 Like that's really boring and being able to think about AI as like just this average transformation like we had SAS a couple of years ago. That is really missing the point of how companies have the potential to transform. So the companies that are going to be able to transform are companies that understand this urgency, have the right people of course and that have that context and are able to think out of the box. First principles I would say is another characteristic that employees have to have and reimagine themselves. And this is where you are going to have a ton of opportunity. But first as in any technological transition it always starts in the same way by value destruction. In any technological change, you first start by seeing who is naked, the losers. And we saw that with the SAS apocalypse, I think in a very very violent way. And I think that the SAS apocalypse is not over even though we have like kind of like positive impressions at one point. I think there is still a lot of value that is going to be destroyed mostly in private equity per my view because public markets have been hit quite a lot already. And then over the next quarters, maybe over the next years, we are going to see these AI winners, the companies that are going to become what I call AGNT.

37:21 Imagine this metaphorical ticker of a company who has succeeded in this new aogentic world. For those of you who are interested, I will be giving away the book that I just published. So, sorry for the second of advertising, but just after this panel, I will be giving it away. And I have open sourced the framework so that everyone can start thinking and reasoning around who is Agent T and how do I become Agent thanks to all of the solutions that you are building.

37:48 >> brilliant. Thank you. Well, I'm going to ask for a 30 second sort of response and final word from everybody. So, Eve, would you like to to kick us off at the end there? >> On which question? >> Just the where the well the response to that and where where value is going to be in 2035. Where would you put your chips? >> sorry. What about chips? >> Where would you bet on value? The value having moved to over the course of the next 10 years.

38:15 >> Yeah. Well, I still stand by my argument that the value is is where you can like own specifications of your system as a human while which basically constrain what AI is doing in your system. So that's always puts human above AI and AI becomes a tool as you're saying and it's a human job to figure out where the specifications are and if you know where where they are this is the biggest value because otherwise you're going to sell thousands agents and foundational models which going to be chaotic and you don't know if it's going to deliver you what you actually want. So either won that >> yeah my view is like in terms of like who's going to win or lose it's it's the same old like you know companies that are that have the right mindset like you know that act with urgency that have the right people and they're investing in the new technologies like right now it's AI in the future it's going to be like something else like you know like the tech world doesn't sit idle like you know like in two years going to see something you know again like amazing like maybe it's you know maybe it's you know the different types of models so anybody like who's moving with agility is going to like win in the business like I I I I I personally don't think that AI changes like who wins and who loses it's still like you know businesses with the right mindset you know will be able to adopt all the great technologies and win and the others are going to lose. Okay, Gabriel.

39:44 >> yeah, impossible to be kept 10 years out, but I think there might be a hollowing out of the the the modest middle. I think that there's a concentration availability at the top where there are access to disordered amounts of talent or data which tend to compound quite well in in this phase. And at the bottom, I think that some companies have completely seen blown away the friction to create based on an idea and to serve an incredibly local base of customers in ways that I think were out of reach. And so in terms of productivity, I I think of it in three pillars. Do the thing faster, which in capitalism do means do it for less. second pillar is do the thing better. My support team now speaks Japanese and nobody on my support team knows any Japanese. That's just pure increase of NPS right there. And the third pillar, which is fantastic, is like do the thing you weren't doing because nobody had the time or the energy or the skills to do. And I think on that third pillar for small companies, there are some interesting opportunities where we might get better personalized service by companies that actually know our names.

40:45 >> Nico, the last word. >> There's there's good companies and there's bad companies. And I I don't think I don't think you need to be that scientific in in defining the good ones and the bad ones. I mean, good companies when there's opportunities, they seize them. When there's new technologies, they adopt them. they kill things that aren't working, and they lean into things that are and aren't afraid to disrupt themselves and others. And, you know, right now, the the the market is is very hot, and I think it's, you know, it's easy for everyone to to to rise when the when there's this much advancement happening.

41:21 but you know, the market won't be hot forever and and when that happens, I think that we'll see a lot of bad companies struggle and that's that's a good thing because it will it will clear up some room for for those that want to win and those that want to do big things. >> Well, as ever, maybe everything changes and everything stays the same. So, we're we're out of time. just last to say thank you very much to our panelists here. please show your appreciation. Thank you very much for being at Raise. Have a great great trip home.