Full transcript of “The Future of Software is Invisible”. Click a timestamp to open the video at that moment. Need another video? Use the YouTube transcript generator.
0:00 SaaS products are in serious trouble. And no, it's not just because someone vibe coded a rust version of the entire Adobe suite or the reverse engineer anything project. It's because we are rapidly approaching a future where it's not millions of people creating software, it's billions. Is your product ready for a world where billions of people are creating software? Or as I like to call it, invisible software. Now, I'm not saying that everybody is going to want to be a coder. Obviously not. The Walmart greeter doesn't want to have to build an app every time they have a problem. They just want their problem solved and their questions answered and their lives made easier so they can get back to the things that they actually care about. But check this out. So we've got benchmarks measuring agents being able to do tasks like fixing bugs and working in a terminal
0:47 and operating a computer. And at the beginning of these tests back in 2024, they were able to do like maybe one in seven. But on regardless of which benchmark you're following, they're all going up into the right. Month after month, they are improving to the point where we're now doing better than four out of five tasks are accomplished well by these agents. So it's just up into the right. We're getting to the point where we are running out of tests that are hard enough to measure agents. But what does that actually mean? That software developers aren't needed anymore? I have a different take. As the rest of the world catches up to what agents can do for them and agents continue to get more and more powerful, we're going to have agents writing software to solve users' problems
1:33 without them knowing or caring it's custom software written ad hoc that's powering the solutions to their problems. So it's not billions of individuals writing code by hand at the speed of humans, it's billions of people whose agents write, commit, and push code at the speed of AI. Let me show you what I mean. But before we get into that, let's hear a word from our sponsor. Let me tell you about Warp Build. CI is a problem for me. I've got to have validation before I merge a PR or push out to production, and that becomes even more critical the more serious my application gets, but it also takes longer and costs more as my application grows. Not to mention the volume of PRs and deploys climbs rapidly with coding agents, and that's exactly what's happening here. I've got an 11-minute and 46-second
2:19 validation pipeline. That's not even the deploy part, that's just the validation. And yes, I've got several layers of caching and everything, but I still would like it to be faster, especially because the number of PRs I've been shipping causes my agents' PRs to constantly hit concurrency limits, and that's really slowing me down. So, I installed Warp Build. And with a single change to my GitHub Actions files, I was able to get twice the speed at half the cost. It supports all major operating systems and architectures, including macOS, and unlimited concurrency, so your agents aren't waiting in line for the next available runner. And I'm only getting started with Warp Build, so I'm excited to look at optimizing further with them with unlimited caches for artifacts and NPM packages and remote container
3:05 builders. And they've got an MCP server, and you know how much I love MCP servers. Agent-ready products for the win. And goodness gracious, it's even showing me recommendations of how to save even more money by downgrading runners that don't need such a beefy box. That's sick. I'm glad I finally stopped letting my CI stop me from shipping at the speed of AI. If that sounds good to you, click the link in the description to get started with $50 in free credits on Warp Build. My barber's name is Victoria, and she's awesome. Let's imagine that she gets a notification from Grokbot or any other proactive bot. Well, let's say Grokbot, for example, saying, "You have four open slots Thursday. Want me to text regulars who are due for a cut?" Now, Grokbot actually is proactive like this, and so this is a sort of thing that could actually happen
3:51 today. And she says, "Yes, please." And so GrogBot builds an app using Cody in this case to find regulars who are due. And so, the code has been written, committed, pushed, and deployed. And now Victoria has a custom app that's just right for her. Now, today a lot of people's experience is just text, but you're watching this unfold in these agentic applications, these assistants, where they're just becoming more and more useful. And sometimes a visual app is exactly what people are looking for. So, the code is written on demand. Victoria's like, "Oh, great. Let's look at the ones who are due for a cut." And so we're going to send a message to all of them to text them on Thursday. And I'll just add a thanks, and then I will send this to four clients. And this app
4:37 is connected to her iMessage, and it's also connected to her Stripe. And so she's getting those notifications from the Stripe app, and she's seeing it open and live in real time in this custom app that she didn't think to build. She didn't build it. Her agent built it because that was the most effective way to solve this particular problem. And so now she's talking to her friend Max, who is also a barber. And Max sees this app, and she's like, "Oh, man. Like, I want that." So, Max says, "Hey, I want that what Victoria has for her barber shop, but with a book now button. So, I'm going to make a customization." And in the Cody ecosystem, Victoria's app might be a public app, so ChatGPT is able to just to fork it. But whatever it is,
5:23 whatever this future looks like, ChatGPT is perfectly able to use the connections that it has to his Square store, and build out this version of this app. So, it's a fully custom, not only the UI and like the look and feel and the text on the page, but even completely new features. And so he's able to send the text on a custom domain, and and it's "Okay, pick a time." And there are apps that exist for this, but the like, do any of them allow you to have this level of customization? They don't. Do uh the users of these apps really care that much that they don't allow this level of customization? Maybe not, unless they start seeing what everybody else is doing and being like, "Well, I kind of wish that I could customize the text or I could send this
6:09 out or I You know what? I'm not using Stripe. I'm using Square. I kind of wish that I could have that, but in my Square app, but it doesn't have it." You hear that all the time. People are complaining about the limitations of their experience. And more and more people are going to be bringing those complaints to agents. And they'll just get used to the idea that if they have a problem, they just take it to the agent, and then the agent writes software to solve that problem. And the person who's using the agent, they don't know anything about software. They don't have to, because the agent is so good at writing software. And I want to remind you, like, this isn't just like, "Wow, this is going to be an exciting future." This is really important for you as a software practitioner, as a builder of systems, to understand and know. So, we're going to talk about that here in a
6:54 bit. All right. So, Max has his thing. Now, Victoria, um in inside of Cody, has said, "Hey, if Max ever changes anything in his version, I want you to let me know." And so, now she sees, "Oh, he's got this book now link. That's awesome." And so, um GrokBot says, "Hey, uh just so you know, he added this. Would you like that?" And so, now she says, "Yes." And boom, she has a fully custom version of her app, but it has now this feature. And this is again the sort of thing that many apps that service these types of individuals, like scheduling apps or whatever it is, um like a a content management app or whatever, they don't have this level of customization. And it's kind of important that they do, or that they do
7:40 something else to back these services um because like they're still using stripe here. But I having a service that can plug into the custom apps that your users are ultimately going to be creating. Now, they're not going to say, "Hey agent, build me an app." They might, but most of the time they're just going to say, "Hey agent, I've got this problem." The agent will proactively build an app because that's the best solution for what they're trying to do. And so now Victoria has her version of the app that runs with stripe and it integrates with all of her stuff and Max has his version of the app that integrates with his Square and everything else and and they're building out these wonderful apps and all of these experiences. Now, some of you might be thinking, "Hey Kent, hold on a second. You're building Cody and Cody is going to be in big trouble when all of these agents are proactively
8:27 building software because they can all do it. You see Chat GPT is building this exact sort of feature right now. They're they're building their sites feature and and you got to believe that Grok and Muse, they're all going to be building their different version of this custom software thing." And so what where does Cody fit in in all that? Like am I going to be ruined? Like no, no. I'm thinking about this as well. This is part of what motivates me building Cody. And so let's say that Max decides, "I'm going to go over to Muse now. I I don't want to use Chat GPT. I'm going to use Muse now." So he says to Muse, "Hey, I want to make a change to my booking scheduler." And Muse is like, "Well, I got bad news for you. I I'd love to and I can, but I can't get to the code. I can rebuild the code, no big deal, but we're going to have to set up your integrations again with Square."
9:14 And Max is like, "Well, I I guess that's easy. Go ahead, but I have like so many apps that I've built on Chat GPT. I guess I'll just stay over at Chat GPT. I don't want to bother with this." And so the walled gardens is a major problem that is the what I am doing about this vision of the future that I have where everybody's building apps. I'm thinking people don't want walled gardens. People want to be able to move, and that's where Cody is positioned. And it's especially useful in a team situation where everybody's using a different agent. You want to have a common place for all of them to go. Now, I'm not here to sell you on Cody. What I'm here to tell you about is that these are the things I'm thinking about with my product. Now, what are you thinking about with your product? How does this affect what you're building? Let's take a look at one interesting example. This
10:00 kind of made the rounds a while ago. This guy, Brandon, wrote coded replacements for seven of Adobe's apps in Rust, and they're all open source. And this like blew up. A lot of people were like, "What the And And it went way beyond just this post." It just went into this idea of like, "Well, software is dead." and whatever. And if you actually used these apps, and some people did, there's a lot of problems with these apps. They didn't It's definitely not done. But, can we just take a step back for a second and say, "Seven of Adobe's Creative Suite apps were rebuilt in Rust by an individual." Wow. Wow. And so, like the problems absolutely can get fixed. And this just
10:45 kind of changes the game entirely. Now, I think that there's still room for software development. And And probably Adobe will will just be fine. Everything will probably be okay. But, I think that this is just a really interesting tell of what the capabilities are here. The fact is, software development's not going anywhere. We still need banks. We still need airlines. We still need grocery stores and internet infrastructure and other institutions. And those places still need software. And it's not just going to be a solo business owner just like telling the agent like, "Go build me this booking service for my airline." Like, that's just not going to happen. You need to have somebody who is in charge of that and is responsible, has ownership, all of that stuff. Now, maybe
11:30 smaller teams can do more, and that's a good thing. I think that there will be more software built by software teams as well as more software built by non-technical people through their agents. That's that's the whole point here. So, not everything is going to be personal software. Don't worry, software engineers are still going to be a thing in the future. But, how you solve problems for users will change shape significantly. You need your software to be agent ready. You need to make it easy for agents to integrate. It need to make it it really easy for like a non-technical user to be able to hook up their system, whatever it is, whether that be in Cody or whether it be in ChatGPT or whatever. Make it easy for them to connect that so that they can have an agent that builds the software that they
12:16 need or answers the questions that they have and all of that. That is going to be the window into your software system because it's just better. It literally is. And so, this is a lot like the birth of the internet. At first, a lot of people were resistant to have a website, and now you'd be crazy not to have a website. And there are actually a couple of good sites to help you get an idea of how agent ready your personal website is. So isitagentready.com from Cloudflare has a good report that will tell you where you're at. And the is-agentic.com site, also from Vercel, has a good way to to measure how agent ready your site is. And if you're building something that requires a lot of documentation or something, then Netlify has a score page
13:02 that you can go to. None of these are sponsored. I just noticed that these are helpful in helping you get your site agent ready, so I thought I'd share those. So, is your app ready for the next billion users? You've got to think about this, and it's not just like, "Oh, yeah, I've got a website and it, you know, agents can click around." That's not efficient. You've got to think about how to make it as helpful and efficient to agents as possible because if you're not as friendly to agents as your competitors, your agents will just go to the competitors. The users don't care. And so you're you need to be attractive to the agents that the users are using. Uh this has implications on scale, but it also has implications on like what the users are actually going to do. I- if you need to think about what does it
13:48 mean that users are not going to come to my website anymore? And they'll just be using endpoints for my website. Is it even valuable in that world? Uh does it kind of change the offering that you have if the users are building custom software with their agents and they don't need to come to your site to use those services. How does that change? You need to think about maybe it's time for a pivot or just some sort of subtle change to make my site ready for billions of users' agents coming to use my services. So I'm just asking you to keep an eye on it. This has been Better With Kent where you and I get better together and I would invite you to like, comment, and subscribe, and share. Thank you so much for getting better with me.
