Transcription
Open Core might have just killed 80% of all apps. And if you're building apps with AI coding tools, you might be worried. People are calling this the year of personal AI agents. And with that comes a dramatic amount of implications for people building apps and software with AI. But in this video, I want to break down the five categories of apps and software that I think will survive so that you know exactly what you should be building with AI coding tools today.
If you don't know me, my name is Chris and for the last 15 years, I've been designing apps and advising startups on product and design. And with that said, let's jump straight into the video.
So let's jump into what has happened over the last few weeks in software and what's happened with open core and why this might be the end of all software or why people are proclaiming it to be the end of all apps and all software and why I don't think that is the case and why I think there is still a ton of opportunity in apps and in software to be building this stuff with AI coding tools still. And I'm going to highlight in this video, if you watch through to the end, the five categories plus a little bonus category of apps and software that I think survive this kind of year of personal AI agents that people are calling it.
Now that Peter has from OpenClaw, the creator of OpenClaw has gone to OpenAI to help them build what will basically be the personal consumer-facing AI agent for everybody to use. I think there is a lot of implications here. So, let's break it down and I'm going to go through the kind of state that we're in now and then cover the categories that you can still continue to build using AI coding tools. And then I'm going to break down exactly what you should be doing next if you are thinking about building something with these coding tools, Claude Code, Codeex, any of these AI kind of vibe coding tools and coding tools and where you should kind of just the things that you should do next basically.
So this is what I'm calling the software bloodbath to start off with. So a few weeks ago, so this is from February 4th, um David Andred posted this. SAS is dead. Agents killed it. Basically, Anthropic released a if you haven't already used it, Anthropic released a product called Claude Co-work a couple of months ago in the Claude desktop app and you can now access that on Windows as well. And what co-work is is it's basically a more user-friendly and less code-focused version of claude code that can do a bunch of work for you. And basically what happened is that anthropic then released a set of plugins for claude co-work which allowed people to do specific jobs like legal uh finance all of these plugins specifically tailored to these specific job roles. And that release of those simple set of plugins within Claude Co-work triggered what is being known on Wall Street as the SAS apocalypse. Basically, all of the software companies, stocks that are from publicly traded software companies just absolutely dive bombed and 285 billion was wiped out in 48 hours. And I think it's now approaching nearly a trillion dollars of value was wiped off the stock market value in just basically a matter of weeks just from these plugins. And then from the release of this next thing that I'm going to talk about here which you probably already know is the reason why you clicked on this video. I'm talking about open claw of claw of course in it Salesforce all down massively and this was just from some simple plugins for claude co-work and they called it the SAS apocalypse.
So, here is the agent that changed it all. Open Claw. So, I've been on paternity leave for the last 4 weeks, which is probably the most crazy time to be not creating content. And Open Claw basically sent a huge amount of shock. Like, it sent shock waves through the entire AI world, has been something that has gone incredibly viral and all because it is basically a personal AI assistant that you can set up that uses your computer. So, people are buying Mac minis, people are buying their own computers to run this thing. So that you're giving an AI agent its own computer, its own device, its own browser, access to all of your applications to basically work on a computer for you. And all you do is just message it in WhatsApp, Slack, or Telegram or any other messaging app that you can connect it to. So I basically text my AI agent in my WhatsApp chat and say, "Can you do this work for me?" It goes away. It uses its own computer, does the job, and then comes back to me. And it also proactively suggests a bunch of other stuff that you can do here as well. They call it the AI that actually does things. Peter Steinberg basically built this incred incredible open- source projects. And yes, there is a lot of security issues with it. There's a lot of bugs with it. It's a very difficult thing to set up. I have got mine up and running just recently coming back from paternity leave after this whole thing went down. And the post I want to draw your attention to here is this post where Peter Steinberger has basically said, "This is the part that will eat so many services going forward. Your interface doesn't matter anymore. Your apps won't be opened anymore. Your users consume content via their existing interface. Context aware only what matters right now. There's an opportunity to rethink the whole operating system."
So this is the idea that your agent is now the operator of your computer and that your agent is the primary operator of basically completing any task for you. Whilst I think this is partially true. So we've got this graphic here of like the agent basically your AI agent saying I am the interface. Now I think the agent and the personal agent will be the interface for a lot of things but I also think that there's still a huge opportunity in software in specific types of software and specific types of applications. And I run a community my community is called the AI founders academy because it's not just about building software with these tools. It's not just about building apps with tools like Claude Code and Co-work. It's about building businesses, building an actual sustainable kind of business and projects that can continue to make money and grow even if it's not necessarily just the app that you're building.
So Claude Claude, what was originally Claudebot and is now OpenClaw is an open source AI. It lives in WhatsApp and Slack and it runs on its own computer. Your basic way to think about this is chat JBT gives you answers and openclaw actually does actions for you from a personal AI assistant point of view. You can think about this as someone who books flights, manages emails, builds tools, but it can do more much much more than that. It can build custom dashboards, custom agents for you. It can run hosts of agents. There are people doing crazy things with openclaw already and all just interacting with it through basically your WhatsApp or your telegram interface. And this there's a huge like security issue with this. So apparently 30,000 exposed instances of open claw. Um there's also people saying like block it now and don't use it. So it's got a lot of hype and a lot of traction and Peter Steinberg has now gone to open AI to partner with them to make the future of personal AI agents. But I think these personal AI agents are more on the consumer side. um personal AI agents will be very very relevant in a generalist sense from a consumer point of view and they will but they will also be very very relevant for companies from a specialist point of view and this is what I'm going to get into talking about in a minute.
So what are the 80% this this 80% of apps that people are claiming that die that are not going to work anymore? Well, I think for a start I think it's not 80%. So I think this actually here is not necessarily 80%, it's going to be more like maybe more like 50%. I think there's still a huge amount of value in software and applications. And I think there's this huge p perception gap basically that people have between how people are interacting with computers right now and how the general population will interact with computers going forwards. It's not going to be the case where suddenly 100% of people switch over to using AI agents as their primary interface to interact with a computer. It's going to take a while. It's going to take a while for that perception to shift. I still have my kind of nan using a computer from like 10 15 years ago and she is absolutely happy using that. these things, this curve of adoption takes time to actually take place. And now this is all accelerated because AI is so much faster than everything else. And maybe we can rethink the entire operating system layer, but there's still a period of time where there's opportunity for people to build applications and software businesses using AI.
And so the 80% that dies is going to be basically thin rappers. So like a nice UI on a database or really basic CRM that don't have any like agent functionality that don't have any way of like interacting with that with an agent basic schedulers or invoicing and also feature only apps as well. So where you think this is a nice one single feature I'm going to build it into an app and try and make some money with it. There's basically no data, no community and no moat around this. So there's nothing there's no nothing significantly valuable about that that will cause people to come back to it and that can create a sustainable business for the long term with that type of app because inevitably the AI models are going to be able to eat that functionality through these personal AI agents where not only can the AI just answer a question for you, it can then go and use that app on your computer. It can go and do that thing for you. It can open Photoshop. it can open whatever tool it needs and then go and do that thing for you. The last one that's going to die here and I do think there are more categories that are going to that are going to potentially fade out in software and applications but is this fast or lifestyle kind of SAS. So this is basically like to-do apps or habit trackers or health apps or anything lifestyle related which is about improving yourself. Your personal AI agent is going to have a so much broader like and deeper set of data towards what you're doing in your day-to-day life. It's going to build up that memory, build up that context over time that these specific habit tracking or health related apps that try and kind of solve those problems for you aren't going to have anywhere near as much data unless you build a really really specific agent that can access all of that that historical data that you have already and then tap into that and draw some kind of like some kind of conclusion or like recommendations from that.
And so what do you what do you build? This is the question. What do you actually build? I think there is still so much opportunity. It just changes what we build. So, open floor and claude co-work and all of these recent tools just changes the things that we build. It doesn't stop us from building anything in the first place. It changes the technology that we build. Technology and software is still going to be a huge move forwards. And I want to draw your attention to this article here from Steven Sonowski who is a former like A16Z um investor I think zed for people in the UK I should just say zed and he wrote a piece called death of software n so too much of what is going on is a race to get to a theoretical end state of a whole new world of business and technology and I think this is the perception this is the way that people think when in this kind of all orno way of thinking about new technology and new innovation is that it's not all it's never all or nothing. We don't always go from this theoretical from this basically like current state of affairs, current state of how we operate to a theoretical end state where everything is the same across a level like it doesn't move to a world where everything is 100% through your personal AI agent. I mean maybe it does but I really don't think that that's going to be the case. Many pushing this around uh were around for the tail end of the transition to an internet-based economy and are struggling with what appears to be a compressed timeline. Much the way people record disruption as a moment versus a journey. So, we're on this path where a lot of different things, a lot of different stages exist at the same time. This creates a lot of leverage and a lot of different perceptions in different groups of people in society. And we're moving to a world where not just one way of existing works. It's a world where the agentic kind of AI also exists alongside a lot of everything else. And sure, some things will die, some things will continue, but it's not all going to be replaced. I've been thinking about where we are today with AI and how the transition to this next wave generation platform of computing might look like previous transitions, but it really isn't. I'm thinking about three different transitions: PC and graphical interface, the shift to online retail, and the pivot to streaming. and he breaks down those in more detail in this post.
AI changes what we build. I want to highlight this specifically. This is the bit we need to focus on. AI changes what we build and who builds it but not how much needs to be built. We need vastly more software not less. And I think that is the key takeaway here. We need more software and the world needs more software for more use cases for more problems to solve more things. And that may be agentic in its nature. that may be surfaced primarily through AI agents, but there is a bunch of software that's tech and technology that still needs to be built using these AI coding tools and a huge amount of opportunity.
So, let's get into the survivors here. So, the first surviving category of software is going to be the data gold mine. So, these are stores of data of very high quality structured data. So think about directories. Think about places where you have a lot of information about a lot of things and a very very high quality that is accessible by agents. So AI agents are hungry. They need to access data for research to be able to do their job to the best possible standard. So we think about building data application companies for agents primarily where people interacting with agents to do their work. People interacting with these agents to actually accomplish tasks will be able to have their agents go and read this data, read these applications, this store of data in order to get the best possible information for their human to make a decision. An example would be like a pricing database across 50 different it says 50 different building material suppliers but like high-quality specific databases for specific things. The more agents there are the more demand there is for structured data and the mo is that these pipelines take time to build and these relationships of people who provide that data also takes time to build as well and I want to go through some examples here as well. So let's think about so a price comparison tool that scrapes and aggregates building materials building material costs across uh UK suppliers. So you can think about being geographically specific here as well. An app that pulls together nursery and child care availability across a local area with weight list times, offstead ratings and fees. A food cost tracker for independent restaurants. So we think about very specific industry niche kind of data sets that agents will want to be able to access.
The second survivor category here is the invisible backbone of technology and software. These are APIs basically. Now the the founder of Zel once said and said in a recent article that one of his thesis, one of his key principles for investing in any com company was that if the company could start as an API first, then they would end up growing into a very big and successful software company. That's paraphrasing. Maybe that's not 100% accurate. Just tell me if I'm wrong on that one. But that's the kind of general gist. Why this is going to be still why this is still going to survive is because APIs are basically the tools the way that agents will be able to do things. They will be able to use APIs to achieve a certain task or to accomplish a certain thing or to get a certain function or something done. So what does the agent call when it needs to do something? It will call an API and the API will be the technology that allows it to do something. Think about fire crawl. It's basically a data scraping API that allows you to scrape any website for the data. Your agent will use the firecrawl API to scrape the data from a website. That is an example of a way that an API would still be relevant. Every agent action is someone's API call. And this is where platforms like Stripe, Twilio, Maps get more valuable because we need these APIs for the agents to go out and read them. So think about niche APIs for local data verification, availability, and some examples of these that might be good would be a booking and availability API for independent hair and beauty salons, for example, that any AI agent or app can plug into to check slots and make appointments. So think about an API layer on top of a set of data or an industry that your agent could then go and access to read to perform a certain task or to check a certain set of information.
Basically the third survivor here of software and applications is going to be these that have the software that has the trust factor and this is basically where you people who have a distribution advantage and an audience creators influencers like myself like a lot of other people and will basically have build applications where those people will buy from the people that they trust. So an application that ties into a creator's specific set of teachings or a set or a specific methodology or an application that basically allows people in a certain audience to kind of gather together. And this is about AI copying. AI can only copy features. It can't copy the relationships between people. And so if you have the ability to leverage distribution and audience as your advantage, you can then turn that kind of that kind of audience and your kind of reason for like existing doing content for people. So the reason I make content is to help people build applications, software and AI businesses using these AI tools and help people find freedom um from doing that and start their own businesses. That is my kind of mission and if I can build AI tools to help people to be able to do that then that is something that people are also going to still want to pay for. So a chef building a recipe app for example, a coach building a fitness app. Even if these are more agentic in nature, like they have an AI agent in them where that coach embeds their specific set of fitness kind of training program to kind of personalize something. So you can almost talk back and forth with your coach as the agent. also creates insane amounts of leverage for those creators and those influencers being able to almost duplicate themselves for anyone in their audience who wants to buy into that ecosystem. And the mo is the brand and the audience because there is a low acquisition cost and people build up a huge amount of trust with you as well.
The fourth survivor here is any application or software with a strong network effect and relationships between people. So the value is in the people not the software. any platform that can build relationships and community. So think about platforms like school that is incredibly valuable and this is a platform that is going to continue to survive because you can't fake relationships between people and you can't build that easily in features. AI can't fake 10,000 people um sharing knowledge between each other. So think about marketplaces, forums, professional networks, niche groups and communities of people and community platforms as well to be able to build communities and groups and relationships on top of and every new user in these platforms makes it more valuable. So think about a vetted only community marketplace for trades people to swap leads or subcontract work, a parents platform specific for specific areas where families share reviews, a niche forum, a niche kind of like job board for AI prompt engineers or in a specific niche. These kind of community platforms are going to be really really strong in even in this kind of per era of personal AI agents.
The fifth survivor here is the niche expert and think about this as vertical AI agent applications. So this is where you build a strong a very very strong AI agent or a set of AI sub aents that have very very deep knowledge in a very specific industry and think about this as an openclaw rapper almost. So openclaw is a generalist personal agent that you can set up in any specific way. Of course, you can make it into a very specialist agent if you want to, but the leverage here and the and the kind of opportunity here is in building your kind of own set of instructions and setup for for something like open claw and then selling that as a package to people, a packaged up agent that does a specific task. And I also think a weird kind of result of this is going to be that a load of people start agent companies with like a name instead of like so instead of like Slack, it would be called like um Sally and you would message Sally and Sally would be really really good at like business management for example. So think about like dental and maybe NHS. I don't think NHS is really relevant here because that's public in the UK. um breweries and HMRC. I think it's these it's these combinations I think which is which is really kind of like pulling together child minders and offstead. I'm not sure unlike these combinations but basically the idea here is that you have a very specifically trained AI agent in a specific industry. You get proprietary data that compounds over time and that agent becomes better and better and better at doing that one specific job.
So, I've got a bonus survivor in here as well, which I think is like kind of kind of like software appy and that can is definitely still going to survive. I want to think I want you to think about this idea of AI agents kind of hiring humans. So, in the current way that we're thinking about it, we think of humans basically controlling the AI agents and that's kind of the only structure that we have. But I think what we're going to go to is this world where humans direct the AI agents and then those AI agents also direct some humans. And you can think about this as agents need to hire humans. And so we need to have a marketplace or some sort of marketplace where agents can go to like find the right humans. And we have marketplaces for this already. But if you think about doing this from an agent first perspective, this could also potentially be an API as well. like a wedding planning agent. If you build like a vertical specific wedding planning agent, the niche expert, then that agent is going to need to hire a florist. It's going to need to hire a B to to make the cake. Um, this is where the human directs agent and the agent directs the human. These are easy to use marketplaces for agents to find human labor. And the moat is the agent visibility of those and the kind of relevance of that marketplace to the specific task that the agent have. And there'll be a bunch of partnerships in here between agent niche vertical agents and these marketplaces. And also the labor availability and quality of those platforms is going to be really really important as well.
So these are the the five kind the six kind of areas that I would continue to build in and there are a bunch more as well. You're not just limited to these six. There is still going to be a bunch more of opportunity in software and applications but these are the kind of ones that I am seeing being really really strong off the back of this kind of era of the personal AI agent that we're seeing with openclaw. And I also just to add another one in there, I also think that AI powered agencies are going to be really really strong as well. So agencies that that deliver a specific outcome for clients that leverage AI tools very very strongly and AI agents and create their own set of AI agents in order to provide a very very strong service for clients and to charge a price that is maybe like 70 to 80% of what you would expect to pay now with a huge amount of leverage because you can do those tasks and get that outcome like 5 to 10 to 20 times faster as an agency. And that basically 20 times the number of clients that you can have that you can do and it increases your profit margins massively as well.
So what do you do next? I think the next thing that we should do here is you should go and build something. Go and build something that you care about. Go and build something that can turn into a business. Think about how AI personal agents are going to impact the the world of software and applications. Do your research, but build something that actually build something that you love, not what not just something that you think you should build based on opportunity. Think about what you love, what your expertise is, what you're really strong in, and then think about how you can turn that into a business, which is incredibly leveraged using AI in this era of personal AI agents. And that is how you're going to win. And start off with the best tools that you can. Claude Code, Codeex, use things like Conductor to kind of run multiple agents building at the same time and learn how to use agent skills as well because agent skills is basically how you turn your AI agents, your AI coding agents into employees almost. I'll be as good at doing a lot of tasks if not better through using skills. And that is what you can do next.
And let's get into my final thoughts. So hopefully you can see that there is still a massive amount of opportunity in the application and software building space but also in the broader AI business building space as well. And that is exactly why I set up the AI founders academy basically to teach people how to build applications, software and also businesses using these AI tools. So you can head over to school.com/iapps if you want to check that out. As always, if you enjoyed this video, don't forget to like and subscribe. I'm going to be creating tons more content on OpenClaw and all of the implications that it has to building things with AI coding tools.