Yeah, so, I mean let's dive a little bit deeper into what makes good products win. Because we actually believe that some of these key data points that we talked about earlier are actually going to determine your success within this upcoming product development era. But let's just start with acknowledging this. Faster coding does not mean better product. Of course faster coding can mean better products, but there is no like straight connection between you building faster and the product becoming better. And we have some data to support this. Pendo, which is a product analytics platform, looked at actual users across thousands of applications and found that only 6% of features drive 80% of all engagement. 6 percent of all features. It's such a small number compared to how much we focus on building new features. CB Insights study companies who succeed and fail, analyzed startup failures and found a 35% fail because they could not have the right market fit. It was not because they couldn't build it, it was because they built something that nobody needed. Looking at Microsoft, I mean one of the tech giants, they actually set up data teams to focus on their experimentation. To see how many of their ideas that they test actually improve the metrics they were designed to move. Two out of three ideas failed to move this metric. And I mean these are not stats that come out like after AI. They have been here all along. So I think most of you who listen to this intuitively know this. But still we see so many organizations that focus on the output rather than the outcomes. They focus on the number of features that's being shipped and this is still a dominant trait that we see. But the organizations who are better at focusing on outcomes rather than the outputs. They usually think a little bit different. They are very good at sitting with this question. Should we build this? Thinking back a little bit, I think actually before AI, teams were forced to ask this question. Resources were scarce. So we had to choose what to build. We could not build all the things that we wanted to. So this created a form of like discipline that it made teams to think before they actually committed on what to build. We really, really wanted to find the things that pushed the needle. But as well as I've been talking about before, AI will change this. When building becomes faster, the old constraint that created this discipline actually decreases. So there's actually a little bit of a risk here that many product teams can lose this discipline on sitting with this question of like, should we actually build this? Is this a good idea? Since we can build more. But again, the teams that succeed are the one that will continuously sit with the question, should we? So how do they do that? How do good product team sit with a question and decide if they should build something or not? Well, the answer is that better products come from doing good product discovery work, putting on the focus on the three pillars that I think most of you know here. Desirability, viability, and feasibility. But Villads was talking about this as well a little bit before that. I think many organizations tend to think that this is something that we do at the start of every project. So when we're going to try to do something new while we go through these different checkboxes, we see if there is a user need. We see if it's viable and we see if we can build it. And then we just build it. But the good product organizations are the ones that does this frequently. At Framna, we have teams that do this exercise almost every week to try to find like the real features to build and how to build them in a good way. So, when we are looking at this model of like desirability, viability, and feasibility, I mean, the first one, the desirability for those who are not that like family with the model, it is about understanding the users and the user context to actually know that we're building something that people will actually appreciate to use. The viability aspect is like understanding the business need and what opportunity that we can unlock by building these kind of things. But then the feasibility question, I mean, the feasibility question has always been like, can we build this? And of course, should we build this? When AI is helping with code generation now, we think actually that this question, well, maybe shrinks or turn into something else. It will change at least. We're not there yet. We do not have the answer, like, what is the right way to ask the feasibility question. We're in the midst of this as well and working on these kind of things. But we still believe that understanding how a feature is implemented, how we will fit the existing system, how it scales, etc. These things will still matter and this will still be valuable questions to sit with before actually deciding what we're going to build. So, AI can help us build faster, but we still need to understand how we can build and sustain it as a whole. I think another big argument why we need to focus even more on doing good discovery work is that we're actually also going through some sort of a platform shift now, we can say. Because we see that our user behaviors are changing now. With the launch of the foundational models, this ChatGPT, Claude, Gemini, etc. We see actually changing user behaviors now. When we go out and do research now, we see users being more comfortable sharing data that we would never think that they would share to anything else. Like, we see them think in different ways of solving their problems. So, for many product companies now, I think it is actually about expanding the research, doing research broader, not only focusing on how does your product solve this problem, but go back to the question, how do the users nowadays solve this problem? Because this might actually be changing now. And again, we will see more products come out in the market markets. We will see more products launch because of that it's, simpler to take an idea to actually like launch something. So, we have to work with this and we have to continuously do this discovery work again and again and again to make sure that our solution or our product is catering to our user needs. But then I think when we start talking about this thing, it's talking about deepening in discovery, I think, well, can AI do this for us? Can we use generative AI to do the research? And here, we want to take a little bit of a stance actually. Good discovery research is about creating new knowledge that doesn't exist. If you go out and do research and make findings that already existed before and that you already know before, then you're not doing discovery research in the right way. AI is actually very good at recycling and transforming like existing knowledge into patterns that you can read and then you can like use. But again, good discovery work is about creating new knowledge. So, we believe that AI can really help you when doing discovery work. And we will come to that a little bit later in the presentation. But we would be skeptical to outsource a lot of discovery work to AI. And one of the persons that we at Framna listen to a lot when it comes to discovery work is Teresa Torres. She's a famous product coach and she's like speaker on the topic of the future of product development. And I think she's sort of put the nail to the head with this one. That she said that we're going to go through a period where companies think that they should build every idea they have. And then we're very quickly going to realize that leads to terrible products. So, coming back a little bit to the numbers that Villads shared, we see that some of the companies that do not spend enough time with problems, do not spend enough time investigating new solutions. We think that this is a place where they might be heading.