The Framework JetBrains Uses to Predict the AI Market Landscape

So with that, let's get started. So with that, let's get started. So we're living in very interesting time So we're living in very interesting time So we're living in very interesting time with AI disrupting development today and with AI disrupting development today and with AI disrupting development today and it's doing so at a very, very high pace. it's doing so at a very, very high pace. it's doing so at a very, very high pace. I feel like new tools and new practices I feel like new tools and new practices I feel like new tools and new practices appear almost every day. appear almost every day. appear almost every day. On the other hand, I know that your team On the other hand, I know that your team On the other hand, I know that your team of market and product researchers are of market and product researchers are of market and product researchers are actually doing a great job of knowing actually doing a great job of knowing actually doing a great job of knowing everything that's going on. everything that's going on. everything that's going on. Um how do you do that? Um how do you do that? Um how do you do that? >> Well, I think by working in this area >> Well, I think by working in this area >> Well, I think by working in this area for almost two years, I noticed the for almost two years, I noticed the for almost two years, I noticed the following. The speed of change is following. The speed of change is following. The speed of change is actually not a big problem as long as actually not a big problem as long as actually not a big problem as long as you can see the bigger picture. you can see the bigger picture. you can see the bigger picture. >> Okay. With speed of change not being a >> Okay. With speed of change not being a >> Okay. With speed of change not being a problem, that's really a refreshing problem, that's really a refreshing problem, that's really a refreshing approach. approach. approach. Um it seems that at first instinct, you Um it seems that at first instinct, you Um it seems that at first instinct, you actually want to spend much more time actually want to spend much more time actually want to spend much more time figuring out all the updates when there figuring out all the updates when there figuring out all the updates when there are so many coming. are so many coming. are so many coming. >> Yeah, that's why I think we definitely >> Yeah, that's why I think we definitely >> Yeah, that's why I think we definitely don't track everything that's happening don't track everything that's happening don't track everything that's happening right now in the market. Instead, we try right now in the market. Instead, we try right now in the market. Instead, we try to understand where everything which is to understand where everything which is to understand where everything which is happening right now comes from and where happening right now comes from and where happening right now comes from and where it's ultimately going. it's ultimately going. it's ultimately going. And this actually helps us to build some And this actually helps us to build some And this actually helps us to build some sort of prism through which we look at sort of prism through which we look at sort of prism through which we look at the world and this prism acts as a the world and this prism acts as a the world and this prism acts as a filter and change us and saves us from filter and change us and saves us from filter and change us and saves us from this change fatigue.

this change fatigue. this change fatigue. >> Cool. What do you see through that >> Cool. What do you see through that >> Cool. What do you see through that prism? prism? prism? >> Well, first of all, we see that AI is >> Well, first of all, we see that AI is >> Well, first of all, we see that AI is already a common thing in the already a common thing in the already a common thing in the developer's life. People know about it, developer's life. People know about it, developer's life. People know about it, use it not only at home with their hobby use it not only at home with their hobby use it not only at home with their hobby projects, but also with their work projects, but also with their work projects, but also with their work projects at the companies including big projects at the companies including big projects at the companies including big ones. ones. ones. >> Okay, so those are the last ones to >> Okay, so those are the last ones to >> Okay, so those are the last ones to adopt something, yeah? adopt something, yeah? adopt something, yeah? >> Well, it's actually depends, but overall >> Well, it's actually depends, but overall >> Well, it's actually depends, but overall we don't question if AI in software we don't question if AI in software we don't question if AI in software development is a thing or not. It's development is a thing or not. It's development is a thing or not. It's already here and likely to stay for a already here and likely to stay for a already here and likely to stay for a long time. long time. long time. Second, we see that agentic coding Second, we see that agentic coding Second, we see that agentic coding becomes a new normal. Developers becomes a new normal. Developers becomes a new normal. Developers starting using more and more of these starting using more and more of these starting using more and more of these agents and they actually not just do agents and they actually not just do agents and they actually not just do this to automate code edits, but they this to automate code edits, but they this to automate code edits, but they actually delegate a whole development actually delegate a whole development actually delegate a whole development task to be executed by the agents end to task to be executed by the agents end to task to be executed by the agents end to end. Hence, this actually changes the end. Hence, this actually changes the end. Hence, this actually changes the whole essence of the development flow whole essence of the development flow whole essence of the development flow from the from the from the regular code validate fixed done in regular code validate fixed done in regular code validate fixed done in editor to task setting execution and editor to task setting execution and editor to task setting execution and review done in in the agentic agentic review done in in the agentic agentic review done in in the agentic agentic chat.

chat. chat. >> Okay, that sounds like a very >> Okay, that sounds like a very >> Okay, that sounds like a very significant change. How did that happen? significant change. How did that happen? significant change. How did that happen? >> Well, there was someone who did kind of >> Well, there was someone who did kind of >> Well, there was someone who did kind of the biggest push last year and that was the biggest push last year and that was the biggest push last year and that was Anthropic Claude Code which by our Anthropic Claude Code which by our Anthropic Claude Code which by our current estimates is used by around current estimates is used by around current estimates is used by around already 8 million of already 8 million of already 8 million of coding professionals and earn a lot coding professionals and earn a lot coding professionals and earn a lot Anthropic around 6 billion in yearly Anthropic around 6 billion in yearly Anthropic around 6 billion in yearly revenue. revenue. revenue. And in general it's considered one of And in general it's considered one of And in general it's considered one of the best the best the best tools for AI coding across all kind of tools for AI coding across all kind of tools for AI coding across all kind of AI tools for the development. And also AI tools for the development. And also AI tools for the development. And also it was the release of Claude Code which it was the release of Claude Code which it was the release of Claude Code which actually started this race of ID actually started this race of ID actually started this race of ID agnostic CLI coding agents. And now we agnostic CLI coding agents. And now we agnostic CLI coding agents. And now we have similar offerings from both like have similar offerings from both like have similar offerings from both like big players like OpenAI and Google and big players like OpenAI and Google and big players like OpenAI and Google and niche companies. niche companies. niche companies. >> Right. So, are you noticing anything >> Right. So, are you noticing anything >> Right. So, are you noticing anything beyond the agentic coding? beyond the agentic coding? beyond the agentic coding? >> Yes, actually we see the same agents are >> Yes, actually we see the same agents are >> Yes, actually we see the same agents are starting moving to the cloud and while starting moving to the cloud and while starting moving to the cloud and while the things like Devin existed for some the things like Devin existed for some the things like Devin existed for some time already on the market, I believe time already on the market, I believe time already on the market, I believe it's just now when this trends actually it's just now when this trends actually it's just now when this trends actually started to mean something. Today with started to mean something. Today with started to mean something. Today with more capable models, with more powerful more capable models, with more powerful more capable models, with more powerful agents and actual developers being agents and actual developers being agents and actual developers being better aware of how to use them.

better aware of how to use them. better aware of how to use them. People start actually to delegate some People start actually to delegate some People start actually to delegate some pieces of work to this cloud agents pieces of work to this cloud agents pieces of work to this cloud agents which sort of like autonomous by design which sort of like autonomous by design which sort of like autonomous by design because they're cloud. because they're cloud. because they're cloud. >> Mhm. >> Mhm. >> Mhm. >> And even though there are some heavy >> And even though there are some heavy >> And even though there are some heavy limitations of what they can limitations of what they can limitations of what they can do, they are already able to solve some do, they are already able to solve some do, they are already able to solve some simple cases or how someone called this simple cases or how someone called this simple cases or how someone called this kind of garbage tasks. So, for example, kind of garbage tasks. So, for example, kind of garbage tasks. So, for example, fixing simple linting errors which fixing simple linting errors which fixing simple linting errors which appeared during the failed CI run. appeared during the failed CI run. appeared during the failed CI run. >> Mhm. So, sounds like we're pretty deep >> Mhm. So, sounds like we're pretty deep >> Mhm. So, sounds like we're pretty deep into all this AI new world. into all this AI new world. into all this AI new world. >> Actually, here is an interesting >> Actually, here is an interesting >> Actually, here is an interesting paradox. Even though kind of almost the paradox. Even though kind of almost the paradox. Even though kind of almost the majority of the developers using AI, AI majority of the developers using AI, AI majority of the developers using AI, AI is not used everywhere. In reality, is not used everywhere. In reality, is not used everywhere. In reality, there are just two activities where AI there are just two activities where AI there are just two activities where AI is used the most. That's coding as is used the most. That's coding as is used the most. That's coding as literal code writing and brainstorming. literal code writing and brainstorming. literal code writing and brainstorming. But we know that development is not only But we know that development is not only But we know that development is not only about the coding and there is a whole about the coding and there is a whole about the coding and there is a whole pieces of software development life pieces of software development life pieces of software development life cycle cycle cycle where AI is not where AI is not where AI is not used at such particular rates. And we used at such particular rates. And we used at such particular rates. And we have a huge room to grow here. So have a huge room to grow here. So have a huge room to grow here. So basically, the dynamics is positive.

basically, the dynamics is positive. basically, the dynamics is positive. The penetration is growing, but not The penetration is growing, but not The penetration is growing, but not equally. equally. equally. >> So how are you tracking these dynamics? >> So how are you tracking these dynamics? >> So how are you tracking these dynamics? >> Well, as you know, at our research team, >> Well, as you know, at our research team, >> Well, as you know, at our research team, we have a dedicated cross-functional we have a dedicated cross-functional we have a dedicated cross-functional group of researchers which does group of researchers which does group of researchers which does exclusively research on all AI exclusively research on all AI exclusively research on all AI initiatives at our company. initiatives at our company. initiatives at our company. And one of our regular projects is a And one of our regular projects is a And one of our regular projects is a pulse, a large quarterly pulse, a large quarterly pulse, a large quarterly run survey about AI run survey about AI run survey about AI that we launch globally in eight that we launch globally in eight that we launch globally in eight national languages across the general national languages across the general national languages across the general population of coding professionals, not population of coding professionals, not population of coding professionals, not just users brains users. We even try to just users brains users. We even try to just users brains users. We even try to hide our identity there to reduce any hide our identity there to reduce any hide our identity there to reduce any sort of bias. sort of bias. sort of bias. And there we collect different data on And there we collect different data on And there we collect different data on the current individual and the current individual and the current individual and organizational adoption of AI and organizational adoption of AI and organizational adoption of AI and this kind of coding tools, project this kind of coding tools, project this kind of coding tools, project metrics like satisfaction metrics like satisfaction metrics like satisfaction and NPS and which also allows us to and NPS and which also allows us to and NPS and which also allows us to track dynamics and do different kind of track dynamics and do different kind of track dynamics and do different kind of analysis. For example, we are able to analysis. For example, we are able to analysis. For example, we are able to estimate the total market size of this estimate the total market size of this estimate the total market size of this kind of tools and how each separate kind of tools and how each separate kind of tools and how each separate player actually earns.

player actually earns. player actually earns. This is our primary source for the This is our primary source for the This is our primary source for the market data. But apart from this, we do market data. But apart from this, we do market data. But apart from this, we do also like more tailor specific research also like more tailor specific research also like more tailor specific research on different kind of topics by request on different kind of topics by request on different kind of topics by request of the product teams which allows us to of the product teams which allows us to of the product teams which allows us to look at the same market from different look at the same market from different look at the same market from different perspectives. perspectives. perspectives. >> Mhm. Okay. So, we've spoken about what's >> Mhm. Okay. So, we've spoken about what's >> Mhm. Okay. So, we've spoken about what's happening right now, but I think for a happening right now, but I think for a happening right now, but I think for a lot of viewers um and for myself as lot of viewers um and for myself as lot of viewers um and for myself as well, it's really interesting to know well, it's really interesting to know well, it's really interesting to know what's the next big thing in AI what's the next big thing in AI what's the next big thing in AI development. So, assuming of course development. So, assuming of course development. So, assuming of course there is still room for another big there is still room for another big there is still room for another big thing. thing. thing. >> Well, of course there is room and there >> Well, of course there is room and there >> Well, of course there is room and there will be something. But before we will be something. But before we will be something. But before we actually jump into this future thing, I actually jump into this future thing, I actually jump into this future thing, I suggest we take a small step back and suggest we take a small step back and suggest we take a small step back and take a look at the past. take a look at the past. take a look at the past. So, So, So, um what was the evolution of AI in um what was the evolution of AI in um what was the evolution of AI in software development so far? So, it's software development so far? So, it's software development so far? So, it's all started with simple implementation all started with simple implementation all started with simple implementation of code completion within the scope of of code completion within the scope of of code completion within the scope of the single line. Let's call this full the single line. Let's call this full the single line. Let's call this full line code completion. line code completion. line code completion. Then with the emergence of the Then with the emergence of the Then with the emergence of the generative AI, we got ability to generative AI, we got ability to generative AI, we got ability to complete code chunks and we got complete code chunks and we got complete code chunks and we got multi-line code completion.

multi-line code completion. multi-line code completion. Then with the the improvement of models Then with the the improvement of models Then with the the improvement of models uh and focus on conversational flow, it uh and focus on conversational flow, it uh and focus on conversational flow, it became reasonable to bring in the whole became reasonable to bring in the whole became reasonable to bring in the whole uh AI chat experience to your IDE or uh AI chat experience to your IDE or uh AI chat experience to your IDE or code editor and we got AI assistance. code editor and we got AI assistance. code editor and we got AI assistance. Then AI features started to expand from Then AI features started to expand from Then AI features started to expand from the single chat to the whole development the single chat to the whole development the single chat to the whole development environment uh linking several parts environment uh linking several parts environment uh linking several parts into a single context and flow. And we into a single context and flow. And we into a single context and flow. And we have seen the emergence of AI code have seen the emergence of AI code have seen the emergence of AI code editors like Cursor or Windswept. editors like Cursor or Windswept. editors like Cursor or Windswept. Then we got agents and we started to Then we got agents and we started to Then we got agents and we started to delegate uh and assign whole end-to-end delegate uh and assign whole end-to-end delegate uh and assign whole end-to-end tasks and a dedicated UI for their tasks and a dedicated UI for their tasks and a dedicated UI for their orchestration, agentic development orchestration, agentic development orchestration, agentic development environments. environments. environments. And then the same agents are moving to And then the same agents are moving to And then the same agents are moving to the cloud and started doing development the cloud and started doing development the cloud and started doing development work more autonomously. work more autonomously. work more autonomously. >> Mhm. So, when you lay it out like that, >> Mhm. So, when you lay it out like that, >> Mhm. So, when you lay it out like that, it feels less like a list of features it feels less like a list of features it feels less like a list of features and more like a trajectory. So, is there and more like a trajectory. So, is there and more like a trajectory. So, is there some bigger pattern behind this? some bigger pattern behind this? some bigger pattern behind this? >> Yeah, let's first draw a line here and >> Yeah, let's first draw a line here and >> Yeah, let's first draw a line here and think. What does this line mean? think. What does this line mean? think. What does this line mean? Why did all these embodiments of AI Why did all these embodiments of AI Why did all these embodiments of AI align in this particular way? And if we align in this particular way? And if we align in this particular way? And if we draw this line further into the future, draw this line further into the future, draw this line further into the future, >> Mhm.

>> Mhm. >> Mhm. >> where will it lead us to? Uh let me actually share our own take on Uh let me actually share our own take on these questions and start with a little these questions and start with a little these questions and start with a little story. So, in early 2025, we were asked story. So, in early 2025, we were asked story. So, in early 2025, we were asked to collect some insights to evaluate our to collect some insights to evaluate our to collect some insights to evaluate our company AI strategy. At the same time, I company AI strategy. At the same time, I company AI strategy. At the same time, I was kind of watching several videos uh was kind of watching several videos uh was kind of watching several videos uh on YouTube on fundamental physics, you on YouTube on fundamental physics, you on YouTube on fundamental physics, you know, like this stuff know, like this stuff know, like this stuff uh like the general relativity, string uh like the general relativity, string uh like the general relativity, string theory, uh standard model of particles, theory, uh standard model of particles, theory, uh standard model of particles, and so on. and so on. and so on. >> Mhm. >> Mhm. >> Mhm. >> And actually, I got quite fascinated by >> And actually, I got quite fascinated by >> And actually, I got quite fascinated by the idea of creating a theory of the idea of creating a theory of the idea of creating a theory of everything, so something that explains everything, so something that explains everything, so something that explains not what is happening at no right now, not what is happening at no right now, not what is happening at no right now, but what is fundamentally possible. but what is fundamentally possible. but what is fundamentally possible. >> Mhm. >> Mhm. >> Mhm. >> So, that's how we ultimately came to our >> So, that's how we ultimately came to our >> So, that's how we ultimately came to our own theory of everything for AI own theory of everything for AI own theory of everything for AI development tooling, uh which we called development tooling, uh which we called development tooling, uh which we called artificial intelligence development artificial intelligence development artificial intelligence development environments framework or AIDE's environments framework or AIDE's environments framework or AIDE's framework. framework. framework. >> So, I can totally see you chilling and >> So, I can totally see you chilling and >> So, I can totally see you chilling and going over fundamental physics after going over fundamental physics after going over fundamental physics after work, but um how do you go from work, but um how do you go from work, but um how do you go from something so big and abstract like this something so big and abstract like this something so big and abstract like this into something that you structure and into something that you structure and into something that you structure and work with uh every day?

work with uh every day? work with uh every day? >> Well, yeah, like any piece of theory, we >> Well, yeah, like any piece of theory, we >> Well, yeah, like any piece of theory, we started with outlining key assumptions started with outlining key assumptions started with outlining key assumptions and definitions in order to abstract and definitions in order to abstract and definitions in order to abstract from the current jargon and sort of like from the current jargon and sort of like from the current jargon and sort of like narrow thinking on the one side, but on narrow thinking on the one side, but on narrow thinking on the one side, but on the other side to kind of lock some the other side to kind of lock some the other side to kind of lock some attributes of reality and make our model attributes of reality and make our model attributes of reality and make our model a little bit uh simpler to use. a little bit uh simpler to use. a little bit uh simpler to use. So, uh the first we introduced the term So, uh the first we introduced the term So, uh the first we introduced the term artificial intelligence system or AIS. artificial intelligence system or AIS. artificial intelligence system or AIS. Basically, this any computer system Basically, this any computer system Basically, this any computer system created by humans that demonstrate the created by humans that demonstrate the created by humans that demonstrate the traits of intelligence while helping traits of intelligence while helping traits of intelligence while helping users achieve their goals or how product users achieve their goals or how product users achieve their goals or how product managers love call this jobs to be done. managers love call this jobs to be done. managers love call this jobs to be done. Here, we basically say that um the Here, we basically say that um the Here, we basically say that um the people want to feel the intelligence people want to feel the intelligence people want to feel the intelligence from the tools, but that doesn't mean from the tools, but that doesn't mean from the tools, but that doesn't mean that this intelligence and this feeling that this intelligence and this feeling that this intelligence and this feeling should be achieved necessarily from should be achieved necessarily from should be achieved necessarily from LLMs. LLMs. LLMs. >> Mhm. So, basically, as long as the >> Mhm. So, basically, as long as the >> Mhm. So, basically, as long as the system helps users achieve their goals, system helps users achieve their goals, system helps users achieve their goals, um the users don't care whether it's um the users don't care whether it's um the users don't care whether it's powered by AI or something else. powered by AI or something else. powered by AI or something else. >> Yeah, for example, take our AI IDs. They >> Yeah, for example, take our AI IDs. They >> Yeah, for example, take our AI IDs. They were always considered as intelligent, were always considered as intelligent, were always considered as intelligent, but the core of our product is not based but the core of our product is not based but the core of our product is not based on the lens, it's based on deterministic on the lens, it's based on deterministic on the lens, it's based on deterministic heuristics.

heuristics. heuristics. So, here we simply say, target this So, here we simply say, target this So, here we simply say, target this experience of intelligence, not try to experience of intelligence, not try to experience of intelligence, not try to put AI everywhere. put AI everywhere. put AI everywhere. Next, we move to artificial intelligence Next, we move to artificial intelligence Next, we move to artificial intelligence development environment or AI IDs. So, development environment or AI IDs. So, development environment or AI IDs. So, basically, this is an AS for creative basically, this is an AS for creative basically, this is an AS for creative software. software. software. Here, we say that there is a huge set of Here, we say that there is a huge set of Here, we say that there is a huge set of tools that are used to create software, tools that are used to create software, tools that are used to create software, which are used at particular stages of which are used at particular stages of which are used at particular stages of software creation process, at particular software creation process, at particular software creation process, at particular levels of work delegation. levels of work delegation. levels of work delegation. Simply speaking, there is a huge world Simply speaking, there is a huge world Simply speaking, there is a huge world outside of IDs, which are opportunities outside of IDs, which are opportunities outside of IDs, which are opportunities which we might not even considered yet. which we might not even considered yet. which we might not even considered yet. Then, we also introduced Then, we also introduced Then, we also introduced common principle agent terms from common principle agent terms from common principle agent terms from economics and sociology we described the economics and sociology we described the economics and sociology we described the relationships between two parties in relationships between two parties in relationships between two parties in case of work delegation. case of work delegation. case of work delegation. Here, we wanted to demonstrate that Here, we wanted to demonstrate that Here, we wanted to demonstrate that agents are actually called agents for a agents are actually called agents for a agents are actually called agents for a reason, and there is a whole piece of reason, and there is a whole piece of reason, and there is a whole piece of theory which describes that. And the key theory which describes that. And the key theory which describes that. And the key point here is that both principle and point here is that both principle and point here is that both principle and agent can be both human or AS.

agent can be both human or AS. agent can be both human or AS. >> Mhm. >> Mhm. >> Mhm. >> And thus, we can consider scenarios >> And thus, we can consider scenarios >> And thus, we can consider scenarios where where where AI principle delegates work to another AI principle delegates work to another AI principle delegates work to another AI agent, and even scenarios where AI AI agent, and even scenarios where AI AI agent, and even scenarios where AI principle delegates work to human principle delegates work to human principle delegates work to human agents. agents. agents. >> Okay, this makes perfect sense in >> Okay, this makes perfect sense in >> Okay, this makes perfect sense in theory, but is it a bit scary? theory, but is it a bit scary? theory, but is it a bit scary? Are we there yet to have AI act as a Are we there yet to have AI act as a Are we there yet to have AI act as a principle and human being an agent? principle and human being an agent? principle and human being an agent? >> Well, let's take a simple example of >> Well, let's take a simple example of >> Well, let's take a simple example of requesting a taxi. requesting a taxi. requesting a taxi. When you request a taxi from Uber, do When you request a taxi from Uber, do When you request a taxi from Uber, do you request it from particular driver or you request it from particular driver or you request it from particular driver or an Uber app? an Uber app? an Uber app? >> The Uber app. >> The Uber app. >> The Uber app. >> All right. And when a taxi driver gets a >> All right. And when a taxi driver gets a >> All right. And when a taxi driver gets a request, your request. request, your request. request, your request. Does they get it from Uber Does they get it from Uber Does they get it from Uber app or you? app or you? app or you? >> It's the Uber app again, so we're there, >> It's the Uber app again, so we're there, >> It's the Uber app again, so we're there, I guess. I guess. I guess. >> Yeah. >> Yeah. >> Yeah. Exactly. Exactly. Exactly. Uh finally, Uh finally, Uh finally, uh we uh started to use terms of uh we uh started to use terms of uh we uh started to use terms of software creation in order to abstract software creation in order to abstract software creation in order to abstract from software development and as a more from software development and as a more from software development and as a more traditional software engineer. traditional software engineer. traditional software engineer. With this, we state the obvious.

With this, we state the obvious. With this, we state the obvious. Uh software developers and technical Uh software developers and technical Uh software developers and technical people alike are only parts of a larger people alike are only parts of a larger people alike are only parts of a larger group who contribute to creation of group who contribute to creation of group who contribute to creation of software. And if these others take the software. And if these others take the software. And if these others take the ownership of the whole process, then ownership of the whole process, then ownership of the whole process, then from their perspective, they sort of from their perspective, they sort of from their perspective, they sort of kind of delegate the development to kind of delegate the development to kind of delegate the development to human engineers. human engineers. human engineers. And the same works vice versa. And the same works vice versa. And the same works vice versa. Uh front-end developers delegate UI Uh front-end developers delegate UI Uh front-end developers delegate UI design to UX designers. design to UX designers. design to UX designers. Back-end engineers delegate cloud Back-end engineers delegate cloud Back-end engineers delegate cloud orchestration to DevOps. orchestration to DevOps. orchestration to DevOps. Engineering teams delegate the promotion Engineering teams delegate the promotion Engineering teams delegate the promotion of the application work to marketing. of the application work to marketing. of the application work to marketing. So, it's all relevant So, it's all relevant So, it's all relevant and kind of depends on perspective. and kind of depends on perspective. and kind of depends on perspective. >> So, it's like Einstein's general >> So, it's like Einstein's general >> So, it's like Einstein's general relativity, but for organizational relativity, but for organizational relativity, but for organizational responsibility, then. responsibility, then. responsibility, then. >> Well, yeah, you can change the laws of >> Well, yeah, you can change the laws of >> Well, yeah, you can change the laws of nature. And if we actually add nature. And if we actually add nature. And if we actually add artificial intelligence here, artificial intelligence here, artificial intelligence here, these relations will likely remain uh these relations will likely remain uh these relations will likely remain uh the same. the same. the same. >> Okay. So, if those were the key >> Okay. So, if those were the key >> Okay. So, if those were the key concepts, what are your assumptions? concepts, what are your assumptions? concepts, what are your assumptions? >> Yeah, so actually, we outlined here uh >> Yeah, so actually, we outlined here uh >> Yeah, so actually, we outlined here uh four key points. First, we assume that four key points. First, we assume that four key points. First, we assume that whatever the future will be, there will whatever the future will be, there will whatever the future will be, there will always be a need to create some always be a need to create some always be a need to create some software, meaning that we won't end up software, meaning that we won't end up software, meaning that we won't end up with situation when we stop using with situation when we stop using with situation when we stop using electronic devices.

electronic devices. electronic devices. Uh digital transformation will continue Uh digital transformation will continue Uh digital transformation will continue and probably we will get even more and probably we will get even more and probably we will get even more software uh in the future. software uh in the future. software uh in the future. Second, the primary driver of change Second, the primary driver of change Second, the primary driver of change will on the market will be the gradual will on the market will be the gradual will on the market will be the gradual delegation of software creation delegation of software creation delegation of software creation activities to artificial systems. activities to artificial systems. activities to artificial systems. Let's be honest, we are all very lazy Let's be honest, we are all very lazy Let's be honest, we are all very lazy and would be really glad to delegate and would be really glad to delegate and would be really glad to delegate some work to someone else, especially if some work to someone else, especially if some work to someone else, especially if we see this work as not valuable. And we see this work as not valuable. And we see this work as not valuable. And actually, the whole human's history actually, the whole human's history actually, the whole human's history supports that, starting with the supports that, starting with the supports that, starting with the division of labor to automations and division of labor to automations and division of labor to automations and digitalization, all of this was done digitalization, all of this was done digitalization, all of this was done just to delegate some work. just to delegate some work. just to delegate some work. Third, artificial intelligence systems Third, artificial intelligence systems Third, artificial intelligence systems will never fully replace humans, who will never fully replace humans, who will never fully replace humans, who will retain for themselves two key will retain for themselves two key will retain for themselves two key activities, task setting and oversight. activities, task setting and oversight. activities, task setting and oversight. Here is speak straight, AI will never Here is speak straight, AI will never Here is speak straight, AI will never replace human developers, but the work replace human developers, but the work replace human developers, but the work will shift towards specifying the intent will shift towards specifying the intent will shift towards specifying the intent and to the reviews. and to the reviews. and to the reviews. >> Okay, those are great news. Thank you >> Okay, those are great news. Thank you >> Okay, those are great news. Thank you for that. AI not replacing humans ever. for that. AI not replacing humans ever. for that. AI not replacing humans ever. >> Yeah, otherwise it's all doesn't make >> Yeah, otherwise it's all doesn't make >> Yeah, otherwise it's all doesn't make any sense.

any sense. any sense. And finally, with high levels of And finally, with high levels of And finally, with high levels of delegation, the personal immersion into delegation, the personal immersion into delegation, the personal immersion into specific development activities will specific development activities will specific development activities will decrease. decrease. decrease. Simply speaking, if you are not the one Simply speaking, if you are not the one Simply speaking, if you are not the one who did the job, you will always be less who did the job, you will always be less who did the job, you will always be less immersed and have less contacts about immersed and have less contacts about immersed and have less contacts about this compared to situation when you did this compared to situation when you did this compared to situation when you did it yourself. it yourself. it yourself. I think this assumption is very I think this assumption is very I think this assumption is very important because it highlights the important because it highlights the important because it highlights the importance of context abstraction and importance of context abstraction and importance of context abstraction and management of uncertainty. management of uncertainty. management of uncertainty. >> Mhm. >> Mhm. >> Mhm. Okay, so it seems like there is a bigger Okay, so it seems like there is a bigger Okay, so it seems like there is a bigger pattern. You spoke about software pattern. You spoke about software pattern. You spoke about software creation process and the stages, but now creation process and the stages, but now creation process and the stages, but now we're really talking about the power we're really talking about the power we're really talking about the power dynamics, like who makes decisions, who dynamics, like who makes decisions, who dynamics, like who makes decisions, who delegates and who does the actual work. delegates and who does the actual work. delegates and who does the actual work. >> Yeah, let me answer you with a meme. >> Yeah, let me answer you with a meme. >> Yeah, let me answer you with a meme. You're absolutely right. You're absolutely right. You're absolutely right. In our framework, we operate within In our framework, we operate within In our framework, we operate within three dimensions. three dimensions. three dimensions. So, the first dimension is stages of So, the first dimension is stages of So, the first dimension is stages of software creation process. Basically, software creation process. Basically, software creation process. Basically, this is our a bit of work definition of this is our a bit of work definition of this is our a bit of work definition of software development life cycle, which software development life cycle, which software development life cycle, which we just focused on the outcomes rather we just focused on the outcomes rather we just focused on the outcomes rather than the process itself. than the process itself. than the process itself. And ultimately here, we got 35 And ultimately here, we got 35 And ultimately here, we got 35 high-level activities grouped into five high-level activities grouped into five high-level activities grouped into five activity groups. From ideation and activity groups. From ideation and activity groups. From ideation and conceptualization to delivery, conceptualization to delivery, conceptualization to delivery, maintenance and feedback collection.

maintenance and feedback collection. maintenance and feedback collection. Let's not stop here because every Let's not stop here because every Let's not stop here because every developer in the room will understand developer in the room will understand developer in the room will understand immediately what this all about. immediately what this all about. immediately what this all about. But what is more novel thing here is the But what is more novel thing here is the But what is more novel thing here is the levels of delegation dimension. So here levels of delegation dimension. So here levels of delegation dimension. So here we define the distribution of roles we define the distribution of roles we define the distribution of roles between the principal and the agent and between the principal and the agent and between the principal and the agent and then attributes of delegation like then attributes of delegation like then attributes of delegation like autonomy level of planning and autonomy level of planning and autonomy level of planning and proactivity. proactivity. proactivity. And the various combinations of roles And the various combinations of roles And the various combinations of roles and values for these attributes define and values for these attributes define and values for these attributes define five levels of delegation. five levels of delegation. five levels of delegation. The first level is L1 tool. Delegation The first level is L1 tool. Delegation The first level is L1 tool. Delegation of very limited and scoped actions. The of very limited and scoped actions. The of very limited and scoped actions. The example here will be a code completion. example here will be a code completion. example here will be a code completion. You just ask AI to finish what you have You just ask AI to finish what you have You just ask AI to finish what you have already started writing. Simple action. already started writing. Simple action. already started writing. Simple action. The next one L2 assistant. Delegation of The next one L2 assistant. Delegation of The next one L2 assistant. Delegation of well-defined sequence of actions or well-defined sequence of actions or well-defined sequence of actions or tasks. tasks. tasks. Which have a very specific boundaries. Which have a very specific boundaries. Which have a very specific boundaries. For example, you want to generate unit For example, you want to generate unit For example, you want to generate unit function unit test for some function. function unit test for some function. function unit test for some function. A simple well-defined task which doesn't A simple well-defined task which doesn't A simple well-defined task which doesn't require lots of context, but it's require lots of context, but it's require lots of context, but it's already a task not a single action.

already a task not a single action. already a task not a single action. Usually we can define here an experience Usually we can define here an experience Usually we can define here an experience of having like a sensation of having a of having like a sensation of having a of having like a sensation of having a third hand. So it's doing some work, but third hand. So it's doing some work, but third hand. So it's doing some work, but it's still kind of your own hand here. it's still kind of your own hand here. it's still kind of your own hand here. The next one is L3 general purpose The next one is L3 general purpose The next one is L3 general purpose executor. This is where we start executor. This is where we start executor. This is where we start shifting from the process to the shifting from the process to the shifting from the process to the deliverables of work. deliverables of work. deliverables of work. L3 agent is able to execute any task L3 agent is able to execute any task L3 agent is able to execute any task which requires some expert knowledge which requires some expert knowledge which requires some expert knowledge input from the principal who acts input from the principal who acts input from the principal who acts more like a consultant for more complex more like a consultant for more complex more like a consultant for more complex topics. topics. topics. The current agent in coding is somewhere The current agent in coding is somewhere The current agent in coding is somewhere here. We believe that like here. We believe that like here. We believe that like state-of-the-art agents like Claude Code state-of-the-art agents like Claude Code state-of-the-art agents like Claude Code and Codex and Codex and Codex already very good at code writing and already very good at code writing and already very good at code writing and low-level solution engineering. But we low-level solution engineering. But we low-level solution engineering. But we still do not trust them in defining of still do not trust them in defining of still do not trust them in defining of what actually should be built because it what actually should be built because it what actually should be built because it requires a deep knowledge of the requires a deep knowledge of the requires a deep knowledge of the business domain. business domain. business domain. And hence, we can fully delegate And hence, we can fully delegate And hence, we can fully delegate execution, but we retain for ourselves execution, but we retain for ourselves execution, but we retain for ourselves task setting and oversight. task setting and oversight. task setting and oversight. The next one is L4 supervised executor.

The next one is L4 supervised executor. The next one is L4 supervised executor. This way we probably move beyond This way we probably move beyond This way we probably move beyond individual scope to organizational individual scope to organizational individual scope to organizational perspective because here the agent perspective because here the agent perspective because here the agent responsible for a specific development responsible for a specific development responsible for a specific development function, for example, managing back end function, for example, managing back end function, for example, managing back end for your full stack web application. for your full stack web application. for your full stack web application. The agent here is supervised because the The agent here is supervised because the The agent here is supervised because the principal role is generally limited to principal role is generally limited to principal role is generally limited to being an approver of key decisions. being an approver of key decisions. being an approver of key decisions. Everything else is Everything else is Everything else is decided by the agent itself. decided by the agent itself. decided by the agent itself. And I think this is also probably the And I think this is also probably the And I think this is also probably the place where you we run out of real world place where you we run out of real world place where you we run out of real world examples. Probably maybe for some real examples. Probably maybe for some real examples. Probably maybe for some real like isolated cases of web coding like isolated cases of web coding like isolated cases of web coding platforms like platforms like platforms like Replit and Lovable. Replit and Lovable. Replit and Lovable. And the final one is L5 And the final one is L5 And the final one is L5 computer center. computer center. computer center. Imagine the following. Imagine the following. Imagine the following. You are a CEO of a startup and you have You are a CEO of a startup and you have You are a CEO of a startup and you have an engineering team by your side. an engineering team by your side. an engineering team by your side. You define what you want to achieve as a You define what you want to achieve as a You define what you want to achieve as a company, company, company, how you do this, what are the key how you do this, what are the key how you do this, what are the key metrics, and whether you're performing metrics, and whether you're performing metrics, and whether you're performing well. well. well. Your engineering team here is just to Your engineering team here is just to Your engineering team here is just to execute your strategy and make your execute your strategy and make your execute your strategy and make your vision a reality.

vision a reality. vision a reality. And actually, you don't care what stack And actually, you don't care what stack And actually, you don't care what stack they will use, what will be the API they will use, what will be the API they will use, what will be the API structure of the application, and where structure of the application, and where structure of the application, and where it will be hosted in Azure VMs or in it will be hosted in Azure VMs or in it will be hosted in Azure VMs or in Docker, in managed Kubernetes in Google Docker, in managed Kubernetes in Google Docker, in managed Kubernetes in Google Cloud. You just delegate all this kind Cloud. You just delegate all this kind Cloud. You just delegate all this kind of decision-making to your engineering of decision-making to your engineering of decision-making to your engineering team. team. team. This kind of delegation is what L5 is, This kind of delegation is what L5 is, This kind of delegation is what L5 is, and trust me, AI is too far from that at and trust me, AI is too far from that at and trust me, AI is too far from that at this moment. this moment. this moment. >> Mhm. >> Mhm. >> Mhm. So, with these levels, you've got it all So, with these levels, you've got it all So, with these levels, you've got it all covered all the way from cute auto covered all the way from cute auto covered all the way from cute auto complete to running an engineering complete to running an engineering complete to running an engineering department on your behalf. Uh but I department on your behalf. Uh but I department on your behalf. Uh but I don't think you said anything about the don't think you said anything about the don't think you said anything about the capability of AI, like how smart it can capability of AI, like how smart it can capability of AI, like how smart it can be. Did you omit that for a reason? be. Did you omit that for a reason? be. Did you omit that for a reason? >> Yeah, that's right. I think there are at >> Yeah, that's right. I think there are at >> Yeah, that's right. I think there are at least two reasons. Uh, first, we think least two reasons. Uh, first, we think least two reasons. Uh, first, we think that capability of AI doesn't mean that capability of AI doesn't mean that capability of AI doesn't mean higher higher higher delegation. delegation. delegation. That's why see why that's why we see all That's why see why that's why we see all That's why see why that's why we see all these fantastic benchmarks which these fantastic benchmarks which these fantastic benchmarks which actually show no limited economic actually show no limited economic actually show no limited economic impact. impact. impact. Uh, besides it, we believe that a Uh, besides it, we believe that a Uh, besides it, we believe that a deterministic system which operates and deterministic system which operates and deterministic system which operates and orchestrates a set of less powerful and orchestrates a set of less powerful and orchestrates a set of less powerful and less capable agents can achieve the high less capable agents can achieve the high less capable agents can achieve the high level of delegation than one super smart level of delegation than one super smart level of delegation than one super smart HI.

HI. HI. Second, everything which I have Second, everything which I have Second, everything which I have described is actually not the attributes described is actually not the attributes described is actually not the attributes of the agent, but the attributes of the of the agent, but the attributes of the of the agent, but the attributes of the relations between the principal of the relations between the principal of the relations between the principal of the and the agent, which is delegation. and the agent, which is delegation. and the agent, which is delegation. And these levels are actually a decision And these levels are actually a decision And these levels are actually a decision of the principal based on their own of the principal based on their own of the principal based on their own personal perception of the agent. For personal perception of the agent. For personal perception of the agent. For example, I might have a task which I example, I might have a task which I example, I might have a task which I want to delegate to Juni at L3 level and want to delegate to Juni at L3 level and want to delegate to Juni at L3 level and allow it to execute like end to end allow it to execute like end to end allow it to execute like end to end autonomously. autonomously. autonomously. But in some more critical cases, I will But in some more critical cases, I will But in some more critical cases, I will kind of switch to L2 level, put Juni off kind of switch to L2 level, put Juni off kind of switch to L2 level, put Juni off sort of leash, and will start feeding it sort of leash, and will start feeding it sort of leash, and will start feeding it some specific tasks. some specific tasks. some specific tasks. Uh, very narrow, very specific. So, even Uh, very narrow, very specific. So, even Uh, very narrow, very specific. So, even if Juni is sort of like technically if Juni is sort of like technically if Juni is sort of like technically capable of working at L3 level, there capable of working at L3 level, there capable of working at L3 level, there will be scenarios where I just want to will be scenarios where I just want to will be scenarios where I just want to delegate less. Hence, the level of delegate less. Hence, the level of delegate less. Hence, the level of delegation is not the attribute of the delegation is not the attribute of the delegation is not the attribute of the Juni, but the attribute of our agentic Juni, but the attribute of our agentic Juni, but the attribute of our agentic contract, which is sort of decided by contract, which is sort of decided by contract, which is sort of decided by myself. myself. myself. >> Mhm. That's really interesting. So, even >> Mhm. That's really interesting. So, even >> Mhm. That's really interesting. So, even though AI is technically capable of though AI is technically capable of though AI is technically capable of doing the job, it's still the humans who doing the job, it's still the humans who doing the job, it's still the humans who decide how much control we're really decide how much control we're really decide how much control we're really prepared to let go of.

prepared to let go of. prepared to let go of. Um, and what's the last third dimension? Um, and what's the last third dimension? Um, and what's the last third dimension? >> Yeah, I think this one we spent less >> Yeah, I think this one we spent less >> Yeah, I think this one we spent less time thinking through. time thinking through. time thinking through. Uh, but I believe it's will be more Uh, but I believe it's will be more Uh, but I believe it's will be more relevance relevance relevance right now. So, this is our right now. So, this is our right now. So, this is our organizational context of the organizational context of the organizational context of the development. Basically, we differentiate development. Basically, we differentiate development. Basically, we differentiate between individual development, SME between individual development, SME between individual development, SME development, and enterprise development, development, and enterprise development, development, and enterprise development, which define different types of which define different types of which define different types of constraints and the complexity of the constraints and the complexity of the constraints and the complexity of the organizational dynamics, which later are organizational dynamics, which later are organizational dynamics, which later are reflected in the complexity of the reflected in the complexity of the reflected in the complexity of the development decisions and ultimately to development decisions and ultimately to development decisions and ultimately to the complexity of the final code base. the complexity of the final code base. the complexity of the final code base. I added this dimension probably not to I added this dimension probably not to I added this dimension probably not to forget about this and actually already forget about this and actually already forget about this and actually already today we see that some things become today we see that some things become today we see that some things become more relevant and more important in case more relevant and more important in case more relevant and more important in case of large organizations. of large organizations. of large organizations. So, this is what we call AIDS framework. So, this is what we call AIDS framework. So, this is what we call AIDS framework. And if we get back to our timeline And if we get back to our timeline And if we get back to our timeline picture, we will actually understand picture, we will actually understand picture, we will actually understand that this line is actually the levels of that this line is actually the levels of that this line is actually the levels of delegation dimension. delegation dimension. delegation dimension. But since our model is much more But since our model is much more But since our model is much more complex, we should we should also complex, we should we should also complex, we should we should also consider how AI penetration grows across consider how AI penetration grows across consider how AI penetration grows across the SDLC life cycle and how it's the SDLC life cycle and how it's the SDLC life cycle and how it's different in different organizational different in different organizational different in different organizational organizational context.

organizational context. organizational context. In our AI pulse, we have a dedicated In our AI pulse, we have a dedicated In our AI pulse, we have a dedicated section on that, which ultimately helps section on that, which ultimately helps section on that, which ultimately helps us to build us to build us to build AIDS maps like this one on the screen. AIDS maps like this one on the screen. AIDS maps like this one on the screen. Here we can actually see where ChatGPT Here we can actually see where ChatGPT Here we can actually see where ChatGPT and cloud code are used, at what level and cloud code are used, at what level and cloud code are used, at what level of delegation, whether or not used, and of delegation, whether or not used, and of delegation, whether or not used, and how they compare to each other on all of how they compare to each other on all of how they compare to each other on all of that and all at the same time. that and all at the same time. that and all at the same time. And since we do AI pulse regularly, we And since we do AI pulse regularly, we And since we do AI pulse regularly, we can also see the dynamics and how all can also see the dynamics and how all can also see the dynamics and how all this changes over time. this changes over time. this changes over time. >> Okay, so this looks really impressive. >> Okay, so this looks really impressive. >> Okay, so this looks really impressive. So, now that we have a holistic model So, now that we have a holistic model So, now that we have a holistic model and we've mapped where we've been and and we've mapped where we've been and and we've mapped where we've been and where we are, we're then able to see where we are, we're then able to see where we are, we're then able to see what's coming next, right? what's coming next, right? what's coming next, right? >> Exactly. Actually, our data shows that >> Exactly. Actually, our data shows that >> Exactly. Actually, our data shows that we are just on the middle of this map we are just on the middle of this map we are just on the middle of this map around like L3 general purpose executor around like L3 general purpose executor around like L3 general purpose executor level. With cloud edges only barely level. With cloud edges only barely level. With cloud edges only barely touching the L4 level. touching the L4 level. touching the L4 level. Also, we still see see here a lots of Also, we still see see here a lots of Also, we still see see here a lots of gaps. So, basically, we have a huge room gaps. So, basically, we have a huge room gaps. So, basically, we have a huge room to grow in every direction.

to grow in every direction. to grow in every direction. However, remember that our framework However, remember that our framework However, remember that our framework describes the fundamental forces and describes the fundamental forces and describes the fundamental forces and provides a more formal description of provides a more formal description of provides a more formal description of reality. And that's actually allows us reality. And that's actually allows us reality. And that's actually allows us to think not what will be the next, but to think not what will be the next, but to think not what will be the next, but what will be kind of the end state of what will be kind of the end state of what will be kind of the end state of this picture and imagine the future in this picture and imagine the future in this picture and imagine the future in this kind of final stage. this kind of final stage. this kind of final stage. Uh let's ask ourselves some questions. Uh let's ask ourselves some questions. Uh let's ask ourselves some questions. What does this L5 level actually looks What does this L5 level actually looks What does this L5 level actually looks like? Where we do where we delegate kind like? Where we do where we delegate kind like? Where we do where we delegate kind of development as a competence on the of development as a competence on the of development as a competence on the organizational level to some intelligent organizational level to some intelligent organizational level to some intelligent system? system? system? What will the process of creating What will the process of creating What will the process of creating software look like? Will it still software look like? Will it still software look like? Will it still operate within the same DevOps cycle? operate within the same DevOps cycle? operate within the same DevOps cycle? Or will we have some completely Or will we have some completely Or will we have some completely different flow of software development? different flow of software development? different flow of software development? What will the human developers be What will the human developers be What will the human developers be primarily doing? Remember our primarily doing? Remember our primarily doing? Remember our assumptions. Human will retain task assumptions. Human will retain task assumptions. Human will retain task setting and oversight. setting and oversight. setting and oversight. And what will the very experience of And what will the very experience of And what will the very experience of creating software feel like for creating software feel like for creating software feel like for everybody? everybody? everybody? Honestly, we can reflect on these Honestly, we can reflect on these Honestly, we can reflect on these questions a lot and we can add more questions a lot and we can add more questions a lot and we can add more questions here a lot.

questions here a lot. questions here a lot. But here what we personally believe will But here what we personally believe will But here what we personally believe will be the primary direction. be the primary direction. be the primary direction. So, first of all, we will experience the So, first of all, we will experience the So, first of all, we will experience the rise of abstractions. rise of abstractions. rise of abstractions. Uh the rise of abstraction. Developers Uh the rise of abstraction. Developers Uh the rise of abstraction. Developers will start thinking less about the code will start thinking less about the code will start thinking less about the code itself itself itself and more about actual solution and more about actual solution and more about actual solution engineering like architecture, engineering like architecture, engineering like architecture, composition, data, and user flows, and composition, data, and user flows, and composition, data, and user flows, and business impact. business impact. business impact. And I think this is quite similar to And I think this is quite similar to And I think this is quite similar to what happened to low-level programming. what happened to low-level programming. what happened to low-level programming. Today, most of us think less about how Today, most of us think less about how Today, most of us think less about how our code is actually run on the our code is actually run on the our code is actually run on the hardware, and we operate via some hardware, and we operate via some hardware, and we operate via some abstractions which are embedded into the abstractions which are embedded into the abstractions which are embedded into the programming languages and programming programming languages and programming programming languages and programming frameworks. frameworks. frameworks. So, the same will happen with the code So, the same will happen with the code So, the same will happen with the code in general. in general. in general. >> So, you're saying the focus will shift >> So, you're saying the focus will shift >> So, you're saying the focus will shift from how to build to what and why to from how to build to what and why to from how to build to what and why to build? build? build? >> Yeah, exactly. The good thing is this is >> Yeah, exactly. The good thing is this is >> Yeah, exactly. The good thing is this is not something new. It just will likely not something new. It just will likely not something new. It just will likely become the core parts of the developers become the core parts of the developers become the core parts of the developers work. work. work. Second, we already see case of Second, we already see case of Second, we already see case of vertical integration of AI across the vertical integration of AI across the vertical integration of AI across the software development life cycle. Today, software development life cycle. Today, software development life cycle. Today, AI probably leaves in our IDEs or chat AI probably leaves in our IDEs or chat AI probably leaves in our IDEs or chat applications, which limits heavily its applications, which limits heavily its applications, which limits heavily its context span.

context span. context span. And the system of the future will likely And the system of the future will likely And the system of the future will likely to integrate AI across the whole SDLC to integrate AI across the whole SDLC to integrate AI across the whole SDLC process end to end. process end to end. process end to end. >> Okay, so instead of jumping between >> Okay, so instead of jumping between >> Okay, so instead of jumping between ChatGPT, Claude, or other tools, AI just ChatGPT, Claude, or other tools, AI just ChatGPT, Claude, or other tools, AI just becomes part of big development becomes part of big development becomes part of big development platform. platform. platform. >> Yes, but not exactly only this. >> Yes, but not exactly only this. >> Yes, but not exactly only this. Actually, people will stop treating AI Actually, people will stop treating AI Actually, people will stop treating AI as a tool and start treating it as an as a tool and start treating it as an as a tool and start treating it as an intelligent resource. And I believe this intelligent resource. And I believe this intelligent resource. And I believe this is the third and one of the potentially is the third and one of the potentially is the third and one of the potentially most fascinating and impactful outcomes. most fascinating and impactful outcomes. most fascinating and impactful outcomes. Let's look at the example. Let's look at the example. Let's look at the example. When we look at the production process, When we look at the production process, When we look at the production process, so for example, we a furniture factory so for example, we a furniture factory so for example, we a furniture factory making furniture. making furniture. making furniture. We always use some raw material to We always use some raw material to We always use some raw material to transform it into the final product. For transform it into the final product. For transform it into the final product. For example, we use wooden planks to make a example, we use wooden planks to make a example, we use wooden planks to make a table. table. table. I believe nobody doubts that software I believe nobody doubts that software I believe nobody doubts that software creation is also a production process creation is also a production process creation is also a production process where the end product is where the end product is where the end product is piece of software. piece of software. piece of software. But if this is true, But if this is true, But if this is true, then what is the input raw material for then what is the input raw material for then what is the input raw material for producing software?

producing software? producing software? And actually, this is intelligence. And And actually, this is intelligence. And And actually, this is intelligence. And for quite a while, we humans were the for quite a while, we humans were the for quite a while, we humans were the only supplier of this intelligence. And only supplier of this intelligence. And only supplier of this intelligence. And within organizations, this was perceived within organizations, this was perceived within organizations, this was perceived as like intelligence as a capability. as like intelligence as a capability. as like intelligence as a capability. Because it's connected to the humans. Because it's connected to the humans. Because it's connected to the humans. But But But when LLMs emerged, organizations got when LLMs emerged, organizations got when LLMs emerged, organizations got another source of this intelligence. another source of this intelligence. another source of this intelligence. So, to So, to So, to artificial intelligence. artificial intelligence. artificial intelligence. >> Mhm. >> Mhm. >> Mhm. >> Uh and it can be actually treated as a >> Uh and it can be actually treated as a >> Uh and it can be actually treated as a variable resource, just like any raw variable resource, just like any raw variable resource, just like any raw material. material. material. >> I actually love this analogy. Um so, >> I actually love this analogy. Um so, >> I actually love this analogy. Um so, this shift is from thinking that AI is a this shift is from thinking that AI is a this shift is from thinking that AI is a tool and seeing it more as something tool and seeing it more as something tool and seeing it more as something that we manage or operate with. that we manage or operate with. that we manage or operate with. >> Yes, and when uh I think organization >> Yes, and when uh I think organization >> Yes, and when uh I think organization will understand that AI is not just a will understand that AI is not just a will understand that AI is not just a hammer, but it's a wood furniture hammer, but it's a wood furniture hammer, but it's a wood furniture factory. factory. factory. That's probably where you might That's probably where you might That's probably where you might experience a fundamental shift in how experience a fundamental shift in how experience a fundamental shift in how software development uh is done. software development uh is done. software development uh is done. And uh as a result, we might end up with And uh as a result, we might end up with And uh as a result, we might end up with something like this agentic software something like this agentic software something like this agentic software production platforms. A vertically production platforms. A vertically production platforms. A vertically integrated software production system uh integrated software production system uh integrated software production system uh where AI is working as an input raw where AI is working as an input raw where AI is working as an input raw intelligence resource, and which is just intelligence resource, and which is just intelligence resource, and which is just organized around the essence of every organized around the essence of every organized around the essence of every kind of work, not just software kind of work, not just software kind of work, not just software development. That is intent management, development. That is intent management, development. That is intent management, execution, and governance. Rather than execution, and governance. Rather than execution, and governance. Rather than traditional DevOps uh cycle. And human traditional DevOps uh cycle. And human traditional DevOps uh cycle. And human developers sort of like managing this developers sort of like managing this developers sort of like managing this production system and orchestrating it.

production system and orchestrating it. production system and orchestrating it. >> Okay, this looks really exciting, but at >> Okay, this looks really exciting, but at >> Okay, this looks really exciting, but at the same time, mildly terrifying. the same time, mildly terrifying. the same time, mildly terrifying. Uh what can we personally do to prepare Uh what can we personally do to prepare Uh what can we personally do to prepare for future like that? for future like that? for future like that? >> Well, I should say that our model >> Well, I should say that our model >> Well, I should say that our model doesn't mean doesn't mean doesn't mean uh that kind of this is exactly the uh that kind of this is exactly the uh that kind of this is exactly the future that will come. future that will come. future that will come. Uh basically, our framework tells that Uh basically, our framework tells that Uh basically, our framework tells that it's fundamentally possible due to the it's fundamentally possible due to the it's fundamentally possible due to the existence of some key forces which drive existence of some key forces which drive existence of some key forces which drive the development of the uh software the development of the uh software the development of the uh software market. market. market. Uh but, it doesn't tell you whether we Uh but, it doesn't tell you whether we Uh but, it doesn't tell you whether we particularly land in this particular particularly land in this particular particularly land in this particular version of reality and when. version of reality and when. version of reality and when. On the other side, understanding this On the other side, understanding this On the other side, understanding this exact fundamental forces really well can exact fundamental forces really well can exact fundamental forces really well can be already a great help for you right be already a great help for you right be already a great help for you right now. now. now. So, try to answer to So, try to answer to So, try to answer to Does AI really grow the opportunities of Does AI really grow the opportunities of Does AI really grow the opportunities of development work work delegation? development work work delegation? development work work delegation? Can it act as this intelligence resource Can it act as this intelligence resource Can it act as this intelligence resource and intelligence raw material? and intelligence raw material? and intelligence raw material? Is it possible to ultimately have a Is it possible to ultimately have a Is it possible to ultimately have a hybrid teams of human developers and AI hybrid teams of human developers and AI hybrid teams of human developers and AI workers? Or is there anything else going workers? Or is there anything else going workers? Or is there anything else going on?

on? on? Then build your own model or use our 80s Then build your own model or use our 80s Then build your own model or use our 80s framework, framework, framework, accept it and start asking another accept it and start asking another accept it and start asking another questions, but within the boundaries of questions, but within the boundaries of questions, but within the boundaries of this conceptual model. this conceptual model. this conceptual model. What will my responsibilities as a What will my responsibilities as a What will my responsibilities as a developer look like? developer look like? developer look like? What skills will be more relevant? If we What skills will be more relevant? If we What skills will be more relevant? If we are speaking about management of are speaking about management of are speaking about management of delegation, does it mean that certain delegation, does it mean that certain delegation, does it mean that certain managerial skills will be a must for managerial skills will be a must for managerial skills will be a must for every developer in the future? every developer in the future? every developer in the future? And if we focus on pure engineering, And if we focus on pure engineering, And if we focus on pure engineering, does it mean that everybody will become does it mean that everybody will become does it mean that everybody will become software architects? software architects? software architects? But then we should have something like But then we should have something like But then we should have something like senior architects, senior architects, senior architects, middle architects, and then junior middle architects, and then junior middle architects, and then junior architects? architects? architects? What skills do I already have? What What skills do I already have? What What skills do I already have? What skills will I need to develop? And am I skills will I need to develop? And am I skills will I need to develop? And am I really ready to change? really ready to change? really ready to change? And then just observe the world, And then just observe the world, And then just observe the world, validate your model and assumptions. The validate your model and assumptions. The validate your model and assumptions. The good thing that change does not happen good thing that change does not happen good thing that change does not happen overnight. And uh yeah, there is always overnight. And uh yeah, there is always overnight. And uh yeah, there is always uh continuous process, there is always uh continuous process, there is always uh continuous process, there is always some friction and some uh inertia.

some friction and some uh inertia. some friction and some uh inertia. And if we have here any engineering And if we have here any engineering And if we have here any engineering leaders watching us now, I think your leaders watching us now, I think your leaders watching us now, I think your objective is much more complex because objective is much more complex because objective is much more complex because you have to rethink many things from how you have to rethink many things from how you have to rethink many things from how the development process uh is structured the development process uh is structured the development process uh is structured to how you can treat AI as this uh to how you can treat AI as this uh to how you can treat AI as this uh intelligence resource, especially how intelligence resource, especially how intelligence resource, especially how you should embed it into your systems you should embed it into your systems you should embed it into your systems where you already have intelligence as a where you already have intelligence as a where you already have intelligence as a human capability. human capability. human capability. But this I believe can be a topic for But this I believe can be a topic for But this I believe can be a topic for another discussion because we already another discussion because we already another discussion because we already have here uh as some vision as a have here uh as some vision as a have here uh as some vision as a company. company. company. Or in the end after all, you can just go Or in the end after all, you can just go Or in the end after all, you can just go to YouTube, watch some videos on to YouTube, watch some videos on to YouTube, watch some videos on fundamental physics, and get inspired as fundamental physics, and get inspired as fundamental physics, and get inspired as we did.