The Scene
The new, well-funded AI startup called /dev/agents is so ambitious that, if it weren’t founded by some of the most prominent pioneers of the mobile revolution, it may not be taken seriously.
David Singleton, Hugo Barra, Ficus Kirkpatrick, and Nicholas Jitkoff, all of whom had front row seats to the biggest platform shift in tech history as alumni of Google, Meta and Stripe, say they’re ready to do it again, as AI ushers in another, potentially more transformative change.
The idea behind /dev/agents, which recently raised $56 million in a seed round led by Index Ventures and Alphabet’s CapitalG, is that the mobile app is all but obsolete, about to be replaced by AI agents that will carry out our bidding without a single tap, pinch or grab.
And they will automatically spring into action when needed. Say you want to send someone a calendar invite for a meeting: One AI agent might focus on managing your calendar. Another might call a restaurant to make a reservation. Another might make sure you have a ride share ready to take you there.
But who will build these agents and how will they work with one another? That’s the problem /dev/agents is aiming to solve by creating a single platform and standard that allows developers to build things they know will work seamlessly with other agents.
Semafor spoke with Singleton and Barra about their plan to upend the mobile world — again — and what it will mean for developers, users and the incumbent walled gardens that build the current system.
The View From David Singleton and Hugo Barra
This conversation has been edited for length and clarity.
Reed Albergotti: Today, you could have an agent on your own network, within your own corporate ecosystem. But at some point, they have to make the leap outside of that, and then it becomes really powerful — where my agent is talking to someone else’s agent to schedule a meeting. Is that where you guys are trying to develop the ecosystem?
David Singleton: If you think about it from a consumer-facing perspective, today, our devices just make us work way too hard to accomplish anything that we care about. Don’t get me wrong, smartphones are awesome. We helped invent a lot of the key components, and we got a lot out of them. But as we all feel the potential that AI brings, we’re really now starting to feel the limitations of the smartphone computing paradigm. Spend any time with ChatGPT and then go and use Google Home or just your phone, you get really frustrated. Like, why can’t it do that? That’s because app and website workflows are really kind of primitive, they’re very repetitive. They treat every user the same way, generically, not individually.
You can imagine today a company like DoorDash could build an AI agent where you go into the app and just say, ‘sushi,’ and it takes care of the rest. Then you’re looking at a screen, you’re still navigating the screen of apps.
Instead of going to the grid of apps to find the app for me, the agents should come to me instead. We think that means that agentic experiences will, for many workflows, actually replace apps today. And what we need is a system that is able to connect me and my personal context to all of the agents that could serve it. And that’s what we’re building.
What does that mean for today’s Android and today’s iOS?
David: What we’re going to build is a cloud-based operating system, so it’s going to work across all of my devices — my phone, my computer, my car. It’ll sit on top of today’s existing set of operating systems.
If I’m Apple or Alphabet, I’m thinking, ‘if this works, then why does anybody need my phone?’ It’s a cloud-based system, I just need some way to connect to the network.
David: It would be very premature to predict exactly how this all plays out, but we do think that there will be quite a significant paradigm shift here in how people use technology. And we think that this operating system that actually sits on top of all the devices is necessary to unlock a lot of the value for consumers. There are a lot of parts of what has been invented with the smartphone and desktop computing paradigm that definitely aren’t going away. Every time we see a technology shift, there are usually very important parts of what’s come before that remain. But new things that you don’t even think of today, like the operating system or the key thing in the UI, tend to become pretty important.
Hugo Barra: Every major competing paradigm shift usually comes in three very large buckets. One is device form factor. The other one is UI, UX paradigm. And then the third one is developer paradigm, typically to do with developer languages getting higher levels of abstraction. The user paradigm stuff has to do with UI, like going from punch cards and the very first computing paradigm of bank frames to multi-touch gestures in the mobile paradigm. Device form factors have to do with the physical thing that you use.
This new computing paradigm that we think we’re staring at the gates of probably will not bring a massive new device form factor, at least not for a while, because the smartphone is just too great. It will come with the second and the third bucket.
The second bucket has to do with UI and UX. The way that apps present themselves today is kind of one-size-fits-all. It’s a very static presentation layer, but we think that this agentic world will bring about dynamic UIs that are task and user-specific. Gen UI is how we’ve been referring to it, as cheesy as that may sound. The other bucket are the developer SDKs, which require complete rethink. We need a React-equivalent moment in the developer world.
You’re saying we’re going to disrupt this duopoly that controls smartphones. But really, they’re the entryway to everything we do online these days?
Hugo: It’s hard to say what will happen in 10 years. The paradigms that are here today are going to change. And the operating systems that exist today are not suitable for this new paradigm that we’re entering. So they either will have to radically evolve, or they’ll be left behind.
David: A good analogy here is the web browser. Before the web, there were various different desktop operating systems, and everyone targeted their app development for the native APIs of each of those platforms.
Then the web came along. We realized we can have an application, the web browser, that actually sits on top of that and is a better fit for all of the things that need to happen in this interconnected internet age.
We imagine something similar happening here. It’ll just feel like a different way of working with technology. It means that for the first time, people should be able to work with technology in much the same way that people work with people today. Building on that involves things like solving the generative UI problem. The right interface to present depends on you, what you’re doing, and the device that you are using right now. If you’re walking down the street or you’re driving, you want an audio-only interface, and we are working so that our system will allow the same agents to present the choices or information that folks need in order to make decisions, to use them across that broad gamut of experiences.
With Android and Apple iOS, drawing developers into that ecosystem was a huge effort. That became this really competitive moat. On the back end, these chatbots are writing code, they’re doing Python queries. Do you think of that differently, in enticing developers to come into this ecosystem that you’re building?
David: Developers are super important. Every shift in technology comes with new platform-level innovation. We have not seen that yet in AI, and it’s that platform-level innovation that enables developers to take advantage of the technology. We’ve seen profound and accelerating innovation at the LLM and foundation layer in the ecosystem. But no platform and distribution ecosystem exists for developers to build on this at scale. That means that the gap between where AI capabilities are heading and developers’ ability to harness it is actually widening.
We talk to lots of developers, and what we’ve heard from them is that it’s quite easy to build demos. So you can use the APIs from the foundation model providers and throw something together that gives you a little bit of a taste of what you could do. But it’s very difficult to go from the demo into production, the gap is huge. Even if you could build some of this stuff, it’s kind of locked up inside your own app or website. So we are building a platform that has three things.
One, that new UI paradigm that we talked about. Two, the right developer SDKs and tool chain to make it possible for folks that maybe don’t have gigantic teams or any AI researchers on staff to actually build great experiences here.
And then also the two-sided marketplace, where we’ll help bring consumers into the ecosystem, because they understand that there’s value here, and we connect them to the developers of these third-party AI agents. It is important to build the apps and the OS together because that means that you get all of the OS services right. It means you get the developer experience right. It also is important to do that so that you have a good density of the right experience as you go out of the gate.
What is an app in this ecosystem? As you were saying, ‘I don’t really want to open up this DoorDash app.’
David: The apps are agents, they will come to you when you need them. We will build the capability with the first-party experience to actually understand your context and then bring the right agent to you. And that’s just a very different way of structuring this ecosystem. It means that there are opportunities for developers to build quite niche experiences. As long as they can have the context that this could be useful for this person right now, we can actually have those quality experiences get in front of users and rise to the top. And that’s very difficult in today’s ecosystem. Unless you’re in the top chart of apps, you’re unlikely to get a lot of discovery. So we’re pretty excited about how this changes the game for developers.
The user doesn’t even know which app they’re using, in a sense, right?
David: It’s still important that there is an opportunity for developers to express their brand and their own personality. It’s great for people to build affinity. That’s part of what we have to figure out: How to make sure that you still know who it is that you’re dealing with because that matters, too.
I have a choice of multiple apps that I could use to call a car today and I know that some of them are much better at dealing with late cancellations and giving me refunds, and others are not. I do think that consumers ultimately care about understanding a little bit of what’s going on and who the players are. We’re working to make sure that we land that in the right way for consumers.
Every time I get an Uber or Lyft now, I just open both apps up and put in the destination, and see which one’s cheaper.
David: That’s a great example of where an agent OS can be valuable. Maybe not even which one’s cheaper, but which one’s actually going to come first. It’s a pain in the ass to have to navigate around.
There’s probably a million different ways to make money in this. How do you see it playing out?
David: We don’t want to share a tremendous amount of detail because obviously, it’s early, and there are different paths that we will explore. There is a thriving business model here at scale. We fundamentally think that we can make a lot of things that matter to people and make it a lot easier through technology provided by AI agents, delivering a ton of value for people. When value is exchanged, money is usually exchanged to pay for it. There’s always a really great opportunity to monetize those things at scale. There’s a couple of different models that we know we could employ here, but you should definitely expect to see us iterate in that space.
Android was brought into Google for a reason. These are big, ambitious ideas. How tough will it be to get out in front of everybody and build that moat? You have some very big, well-resourced companies that probably want to own this space as well.
David: This space is exciting. There’s a real opportunity to serve people in new and different and better ways. There are plenty of companies who are trying to solve this problem. We fundamentally believe, from our experience of having built platforms like this before, that you do have to actually build the applications and the operating system together. That’s why we’ve chosen to take on this scope of a problem. We also believe that this is the team that can execute this vision very well. There’s a real value to being small enough to be very nimble. Again, we learned that in the early days of Android, and I think we have built, so far, a world test team to do this.
On the back end, we have SDKs, we have APIs. Do we have to reimagine how those are built in the AI world?
David: There’s no standard, accepted developer tool chain for building AI agents. We’ve talked to developers inside little companies and inside huge brands. They’re trying to build AI agents, and they’re telling us that anything that is out there is just not working. Everyone is cobbling it together themselves.
Actually, funny story. We were talking to a developer inside a big, well-known consumer brand, and they were very excited about what we’re doing. They said, ‘Oh, I get it.’ It used to be that front end development was really hard and really fragmented. There were lots of different JavaScript toolkits. It was hard to get started. There was no paved path, and also it was impossible to share code between projects that were all done in different ways. And then Facebook started this project, first of all, to make their own experiences better for their thousands of developers. And then they open-sourced it, called React. React has completely revolutionized front-end development because you can start sharing components between companies, between projects. It’s a different way of thinking about development. But once you get it, it’s incredibly productive. That person said to us, ‘There has been no React moment for AI. I think you guys should do that.’ That is what we’re up to.
What’s also different about this is the probabilistic nature of the technology. How do you ensure that these agents will do what they’re supposed to do and not go off the rails?
David: We’ve personally built a bunch of agents, sat with developers who’ve been building agents. We’ve gotten a pretty good handle on how you can use the core of what we’ll build on our SDK to make sure that you have very dependable performance. This is possible, so we will help solve for that. It is also the case that some of the most sensitive things that you might want to do, we as a system will still make sure that users have the opportunity to say, ‘I want to approve those things.’ There’ll be certain actions where an agent can propose them to the core operating system to say, ‘Let’s present this deterministically to the user.’ We think that by combining those things, we’re going to be able to create great consumer experiences and also great developer productivity, which is important.
As we look at the capabilities of these models, is there a threshold of capability that they need to reach? Or do you think we’re already there?
David: A very core thesis of this company is that we don’t need to train our own foundation model. If you had set out to build a company like this two years ago or even 18 months ago, you would probably have concluded that you needed to. But what we’ve seen is that the rate of progress in the foundation models from the labs is such that it’s no longer necessary. We determine that both from our own experience and also from talking to a lot of AI experts. We’ve been very fortunate to be able to bring a bunch of amazing experts into our company as angel investors who are really our advisors, our brain trust. A good example would be [OpenAI co-founder] Andrej Karpathy.
The core thesis is that the capability of the current models is actually good enough for us to unlock real consumer value. As those models continue to get better, which I believe they will, this system gets even better over time. Even if they get uber capable, you actually still need a system like this. You need the ability for developers out there to be able to engage with it. You need the ability for consumers to have their context understood somewhere so that it can marshal all of these amazing agents that might exist for them. We believe it’s good enough today, and we believe it gets better as those fundamental foundation model providers get better. And even if there is perfect AGI, you still need this platform.