Pod Insights Decoder Why tech is racing to adopt AI coding About this episode This is Casey Newton, founder and editor of the Platformer newsletter and cohost of the Hard Fork podcast. I’ll be guest hosting the next few episodes of Decoder while Nilay is out on parental leave. For the next three weeks, I’ll be talking to leaders in the productivity space about what they’re building, and how they can help us get things done.
My guest today: Michael Truell, the CEO of Anysphere, the maker of automated programming platform Cursor AI. I sat down with Michael to talk about his product and how it works, why coding with AI has seen such incredible adoption, and what the future of automated programming really looks like.
Read the full transcript on The Verge.
Links:
Anysphere, hailed as fastest growing startup ever, raises $900 Million | Bloomberg
AI coding assistant Cursor draws a million users without even trying | Bloomberg
Anthropic rehires AI leaders from Anysphere | The Information
Cursor apologizes for unclear pricing changes that upset users | TechCrunch
OpenAI looked at buying Cursor creator before turning to rival Windsurf | CNBC
Interview with Anysphere CEO Michael Truell about coding with AI | Stratechery
Credits:
Decoder is a production of The Verge and part of the Vox Media Podcast Network.
Our producers are Kate Cox and Nick Statt. Our editor is Ursa Wright.
The Decoder music is by Breakmaster Cylinder.
Learn more about your ad choices. Visit podcastchoices.com/adchoices
Transcript 11,420 words · 11 speakers
0:00 Speaker 00 Real innovation is powered by real people. This is AMEX Corporate. Trusted experts, smarter technology, 24/7 dedicated support from day one to answer questions, help set up policies and approvals, and consult on spend capacity as your business needs evolve. Plus, on-demand resources help level up operations at every stage. That's the trusted service you've come to expect. Built to simplify, designed to scale. This is AMEX Corporate.
0:32 Speaker 01 New and exclusive holiday decor just launched online at The Home Depot. Check out the viral Grand Duchess collection with smart holiday lights you can control from your phone. Easily customize the twinkle across trees, wreaths, and garlands to match your holiday vibe. Create stunning light effects, then switch it up with the tap of your fingers. Even the man in red will be impressed. Shop the holiday decor you want online only at homedepot.com.
1:03 Speaker 02 This episode is brought to you by Michelob Ultra. Good times mean competitions between you and your friends, like seeing who can throw some pigskin the farthest or rush the field the fastest. Winner gets an Ultra. Because it's not just a superior-like beer, it's also giving fans superior access to their favorite sporting events. Superior is worth playing for. Visit michelobultra.com/superioraccess to enter. Enjoy responsibly. Copyright 2026 Anheuser Busch InBev Michelob Ultra Beer, Saint Louis, Missouri.
1:33 Speaker 03 Hello and welcome to Decoder. This is Casey Newton. I'm the founder and editor of the Platformer Newsletter and co-host of the Hartford Podcast. And I'll be guest hosting over the next few episodes of Decoder while Neelai is out on parental leave. Congratulations, Neelai. And I'm very excited for what we have planned for you all. If you've followed my work at all, particularly when I was a reporter at The Verge, you'll know that I'm a total productivity nerd. I love productivity apps, whether it is a to-do, or some kind of collaborative app, or something making use of AI. At their best, I think productivity apps are the way we turn technological advancement into human progress. And also, they're fun. I like trying new software.
2:16 Speaker 03 And every new tool brings with it the hope that this is the one that is finally going to complete the setup of my dreams. Over the years, I've used a lot of these programs, but I rarely get a chance to talk to the people who make them. So for my Decoder episodes, I really wanted to talk to the people behind some of the biggest and most interesting companies in productivity about what they're building and how they can help us get things done. And that brings me to my guest today, Michael Truell, the CEO of AnySphere. You may not have heard of AnySphere, but I bet you've heard the name of its flagship product, Cursor AI. Cursor is an automated programming platform that integrates with generative AI models from Anthropic, OpenAI, and others to help you write code.
2:59 Speaker 03 I guess there I will pause and give my disclosure for the episode, which is that my boyfriend works at Anthropic. But Cursor is built into a standard version of what programmers call an integrated development environment, or IDE, with technology like Cursor Tab, which auto-completes lines of code as you write them. And Cursor has become one of the most popular and fastest-growing AI products in the world. And AnySphere, the company that Michael co-founded just three years ago after graduating from MIT, is now shaping up to be one of the biggest startup success stories of the post-ChatGPT era. So I sat down with Michael to talk about Cursor, how it works, and why coding with AI has seen such incredible adoption.
3:42 Speaker 03 As you'll hear Michael explain, this entire field has changed a lot over the past few years. And now, here in San Francisco, tech executives and employees are regularly telling me about how much they are using and liking Cursor. So, look, a lot of people are worried that AI could take their jobs, and rightly so. I would argue, but you'll hear Michael say that job losses are not going to come from simple advances in tools like the one that he's making. Lots of people in the Bay Area think that super intelligent AI is going to remake the world overnight, making products like Cursor pointless. But Michael believes that change is going to come much more slowly. I also wanted to ask Michael about the phenomenon of vibe coding,
4:25 Speaker 03 which lets amateurs use tools like Cursor to experiment with building software of their own, even if they have never built anything before. That's not Cursor's primary audience, Michael told me, but it is part of this broader shift in programming, and he's convinced that we're only scratching the surface of how much AI can really do here. So, AnySphere CEO, Michael Truell, here we go.
5:03 Speaker 03 Michael Truell, you are the co-founder and CEO of AnySphere, the parent company of Cursor AI. Welcome to Decoder.
5:10 Speaker 04 Thank you for having me.
5:11 Speaker 03 What is Cursor? What does it do? Who is it for?
5:14 Speaker 04 Our intention with Cursor is to be the best way to build software, and in particular, the best way to code with AI. For people who are non-technical, I think that best way to think about Cursor as it exists today is think of like a really souped-up word processor, where the way engineers build software is they're actually doing a lot of writing. They're sitting in something that looks like a word processor, and they're editing millions of lines of logic, so things that don't look like language. And Cursor helps them do that work way more efficiently, especially with AI. And there's kind of two different ways Cursor does that right now. One is Cursor is kind of watching you do your work, and it's trying to predict the next set of things you're going to do within Cursor.
5:57 Speaker 04 So this is the autocomplete form factor, and that can be really souped up in programming compared to writing, because unlike writing, it's actually there are often times when you're programming where the next 20 minutes of your work are entirely predictable. Whereas with writing... It's a little hard to get a sense of what a writer is actually going to put down on the page. There isn't really enough information in the computer to understand the next set of things they're going to do. And then the other way that people work with Cursor is they're kind of increasingly delegating to Cursor like they're working with a pair programmer, working with another human. So they're handing off small tasks to Cursor and having Cursor kind of go and tend on them.
6:31 Speaker 03 Well, we'll dig a little deeper into the product in a moment, but first, let's talk about how all of this started. When you founded AnySphere, you were working on computer-aided design software. How did you get from there to Cursor?
6:44 Speaker 04 My co-founders and I, we come from backgrounds where we've been programming for a while, and we've also been working on AI for almost as long as we've been programming. And so, you know, one of my co-founders, one of us had worked on like recommendation systems in big tech. Another one of us had worked on computer vision research for a long time. You know, another one of us had worked on trying to make machine learning algorithms that could learn from very, very, very little data. You know, another one had worked on like a competitor to Google, using the antecedents or the, you know, the things that came before LLM technology in machine learning, but worked on AI for a long time, had been engineers also for a long time, and loved programming. And in 2021, there
7:27 Speaker 04 were two moments that really excited us. One was using some of the first really useful AI products. Then another was kind of this body of literature that was showing that AI was going to get better, even if we kind of ran out of ideas by making the models bigger and training them on more data. That got us really excited in like, kind of a formula for creating a company, which was you pick an area of knowledge work and you build the best product for that area of knowledge work, like the place where you do your work as AI starts to change. And then hopefully you do that job well and you get lots of people to use your thing. And then if you do, you can see where AI is helping them and you can see where AI is not helping them and where the human just has to correct the AI a bunch or just do work without any AI help. And then you can use that to then make the
8:09 Speaker 04 product better and kind of push the underlying ML technology forward. And then that can maybe get you into a path where, yeah, you can really start to build like the future of knowledge work as this tech gets more mature and be kind of the one to push, push the underlying technology to. So we got kind of interested in that formula for making a company. And the craft that we really loved, the knowledge work that we really loved, which was building things on computers. We actually didn't touch it first. We went and we worked on a different area, which was, as you noted, computer aided design. It was trying to help mechanical engineers, which was a very ill fated decision because none of the four of us are mechanical engineers. And, you know, we had friends who were interested in the area, you know, we had worked on robotics in the past, but it wasn't really our specialty.
8:51 Speaker 04 And it was because it seemed like there were a bunch of other people working on trying to help make programmers more productive as AI got better. But after, you know, six or so months of working on the mechanical engineering side of things, we got pulled back to working on programming. And part of that was just our love for the space. Part of that too was just, it seemed like the people who we thought had the space covered. They were building useful things, but they weren't really pointed in the same direction. And they didn't really seem to be approaching the space with like the requisite ambition. And so, yeah, we decided to to build, you know, the best way to code with AI and that's where a cursor started.
9:24 Speaker 03 I have read that one of the AI tools that you used early on was GitHub Copilot, which came out about a year before ChatGPT. What was your initial reaction to Copilot and how did it influence what you wanted to build?
9:38 Speaker 04 Copilot was awesome. Copilot was a really, really big influence and it was the first product that we used that had AI really at its core that we found useful. One of the sad things to us, you know, is people who'd been working on AI and interested in AI for a while was that it was very much stuff that was just like kind of in the lab or in the toy stage. It felt like for us, the only real way AI had touched our lives as consumers was mostly recommendation systems, right? You know, the news feeds of the world, YouTube algorithms, things like that. And so GitHub Copilot was, yeah, it was the first product where AI was really, really at the core that was that was useful. And so that was a big inspiration and at the time we were considering, you know, should we try to pursue careers in academia?
10:20 Speaker 04 Copilot kind of was this existence proof that, no, actually it was, you know, time to work on these systems out in the real world. And, you know, even back then in 2021, there were some rough edges. There were some places. The product was wrong in really obvious ways and you couldn't completely trust its code output. But it was nonetheless really, really exciting. And another thing to note, too, is apart from being the first useful AI product, it was the most useful new dev tool that we had adopted in a really long time. And we were people that had kind of optimized our setups as programmers and had kind of modded out our text editors and things like that. And we were using this like crazy kind of text editor called Vim at the time.
11:03 Speaker 04 And it was, yeah, not just the first useful AI product that we had used, but also the most useful dev tool we had used in a really long time.
11:09 Speaker 03 That's interesting. So you guys sort of like software. You like using software. You like trying to find software that makes you more productive. I feel like that probably made you well suited to tackle a problem like the one Cursor is trying to solve.
11:22 Speaker 04 Yeah, I think caring about the tools we use was helpful. And I think that there are actually kind of different degrees of that on our co-founding team. One of my co-founders in particular is like the straight out of Central Casting early adopter, you know, is the first one on these new browsers, first one on kind of the new category of everything. A couple of us are a little bit more laggards. And so I think actually kind of having maybe that diversity of opinions has helped us in some of the product assessments we've made.
11:50 Speaker 03 So you described Cursor as kind of like a souped up word processor. Software engineers, I think, would call it an integrated development environment or IDE. And developers have been using IDEs since the 80s. But recently, AI Labs have released tools like OpenAI's Codex or Claude Code that can run directly in a terminal. Why might someone use Cursor over those options?
12:14 Speaker 04 Both of those are really useful tools. The thing we care about being, so I think we start as this IDE, we start as this text editor. And what we really care about getting to is to a world where programming is completely changed. And in particular, you know, a world where you can develop professional grade software, perhaps without even really looking at the code.
12:35 Speaker 04 And it's that that kind of future programming and like changing it from this weird, like you're reading these millions of lines of logic in these like esoteric programming languages to getting you to a world where you can build software. By just specifying the minimal intent necessary to kind of to build the software you want, you know, you can tell the computer the shortest amount of information it needs to really get you and it can fill in all of the gaps. And yeah, programming today is this intensely labor intensive, time intensive thing where to do things that are pretty simple to describe, to get them to actually work and show up on a computer, it takes many thousands of hours and really large teams and lots of work, especially at professional scale. So that's where we want to get to is kind of inventing that new form of programming.
13:19 Speaker 04 I think that that starts as an editor and then that starts to evolve. And so we're already kind of in the midst of that where right now cursor is this place where you can work one-on-one with an agent and you can work with, you know, our tab system. And then increasingly, we're getting you to a world where more and more programming looks like starting to delegate your work to a bunch of helpers in parallel. And there's a product experience to be built for making that, making that great and productive and understanding what all of these parallel helpers are doing for you, being able to intervene in the places where it's helpful, you know, understanding their work when they come back to you at a level that's not having to read every single line of code. Yeah, I think that there's a competitive environment with a bunch of tools that are interested in programming productivity. One of the things that's limiting about just a terminal UI is that you have only so much
14:02 Speaker 04 expressiveness in the terminal and control over the UI. From the very start, we've thought that the solution to automating code, replacing it with something better, is this kind of two-pronged thing where you need to build like the pane of glass where programmers do their work and you need to discover what the work looks like, you need to build the UI. And then you also need to build the underlying technology. And so one, one thing that would distinguish us between, you know, some terminal tools is just the degree of control you have over the UI. Another thing too is we've done a lot of work on the model layer, on improving, going beyond just having things that show up well on, on a demo level. And there's a lot of work on AI products to dial in the speed and the robustness and the accuracy of them.
14:45 Speaker 04 And for us, one important product lever there has been building kind of an ensemble of models that work with the API models to improve their abilities. And so every time you kind of call out to an agent in Cursor, it's like this set of models that some of them are API, some of them are custom. And then also for some form factor, or for some of the features, it's entirely custom, for instance, like the supit autocomplete. And so that's also one thing that has kind of distinguished us from, from other solutions.
15:10 Speaker 03 Yeah, let's talk a bit about these proprietary models. They seem to be fueling a lot of your success. When ChatGPT and the OpenAI API first got released, we saw a lot of startups come out that quickly were dismissed as just wrappers for an API. Right? You're just sort of trying to build something on top of somebody else's API. And Cursor started in a similar way where it was using other folks' APIs in order to to create its product. Since then, you're building on top. Say a bit more about what you're building and how you're hoping it kind of sets you apart from those sort of pure wrapper companies.
15:47 Speaker 04 I think that also, like one asterisk before getting into the model side of things. I think that the wrapper term came from the very start of when people are building AI products, when there was only so much time to kind of make the products a bit deeper. And now I think, you know, we're at a point where there's a ton of product overhang. And so even if you're just building with the API models, I think that there's, and you know, lots of areas are area of, you know, working on software development lifecycle, but in other parallel areas too. I think there are very, very deep products to be built on top of those things. And it's in, you know, and that sounds like the wrapper term for at least some, some areas is a little bit dated. But on the model level, you know, from the very start, we wanted to build a product that, you know, got a lot of people using it.
16:29 Speaker 04 And one of the benefits you get from that scale is you can see where AI is helping people and you can see where AI is not helping people and where it gets corrected. And that's a really, really important input to making AI more useful for people. And so at this point, you know, for instance, with our tab model, which does over a billion model calls per day. This is one of the large language models that writes actually almost some of the most production code in the world. And we're also on our fourth or fifth generation of it. That is trained using product data of seeing where AI is helping people, seeing where it isn't, seeing what in the places where it isn't, trying to predict how it can help humans. And also requires a ton of infrastructure, specialty talent to be able to make those models really good. You know, for instance, one of the people who has worked on those models with us is Jacob,
17:12 Speaker 04 who built actually the kind of GitHub Copilot before GitHub Copilot, which was Tab 9, which was the first kind of programming autocomplete product. He is also one of the people who built one of the first million token context window models, and so has done a lot of work on making models understand more and more and more information. But yeah, specialty talent and specialty infrastructure to do that work. And one of the, in our ambling, kind of windy way to working on Cursor, I think one of the things that really did help us was when we were working on CAD and also in some of our explorations before, my co-founders had to dig very deep into kind of the ML infrastructure and modeling side of things. And so when we actually set out to work on Cursor, we thought it would be a long time before we started to do our own modeling as a product lover, but it happened much sooner than we expected.
17:57 Speaker 03 Recently, I had dinner with the CTO of a big tech company, and I asked him about what coding tools were popular with his engineers. And he told me that he actually regularly surveys them on this question. And they had Cursor available as like a trial, it was labeled. And he said he was getting these panicked messages from engineers saying, please tell us you're not about to take away Cursor, because they've become so dependent on it. Can you give us a sense of why, for programmers, this has kind of felt like a before and after moment in the history of the profession? What is it that tools like Cursor are making so different in the lives of these engineers day to day?
18:35 Speaker 04 I think that we're just already at a point where the, you know, we are far, far, far from the ceiling of where things can go, and far, far, far from a world where, you know, much of coding has been replaced with something better. But, you know, I just already at this point, these products and these models can do a lot for programmers and are already taking on, you know, quite a bit of work. And I think that the technology is especially good for programming for a few reasons. You know, one is that programming is text-based, and that is the modality that the field has figured out perhaps the most. There's a lot of programming data on the internet, too. So a lot of open source code. Programming is also pretty verifiable, too. And so one of the important engines of AI progress has
19:19 Speaker 04 been training models to predict the next word on the internet and making those models bigger. That engine of progress, you know, has largely run its course. There's still To do there. But the next thing that's kind of picked up the torch in making models better has been reinforcement learning. So it's been basically teaching models to play games, kind of similar to how in the mid 2010s, you know, humanity figured out how to make computers really good at playing Go and playing Dota and other video games. We're kind of getting to a level of of language models where they can do tasks and you can set up games for them to get even better at those tasks. And programming is great for that because you can run it, write the code, and then you can run it. And then you can see the output and see if it's actually what you want. And
20:03 Speaker 04 so I think there's a lot about the technology that makes it especially good for programming. And yeah, it's just, you know, I think one of the use cases that's the furthest ahead in kind of deploying this tech out to the world and people finding real value from it.
20:14 Speaker 03 Yeah. I mean, my sense is maybe if I used to have to work eight hours a day, now it's maybe closer to five or six. Is that part of it?
20:21 Speaker 04 Yes, in the sense that I think that the productivity gains of what would have taken you eight hours before in some companies now actually can take you five or six hours. I think that that is real, not across all companies, but is really real in some companies. But I think that the thing I would nitpick on there is I don't think programmers are actually just working, you know, are shortening the hours that they're working. And I think a lot of that is because there is just a ton of elasticity with software. And I think it's really easy for people who are non-technical or just don't program professionally to underrate how inefficient programming is at a professional scale. And a lot of that is because programming's kind of invisible. You know, what a programmer is doing at, you know, a company like Salesforce
21:06 Speaker 04 is there are just tens of millions of lines, many millions of files of existing logic that describes how their software works. And anytime they have to make a change to that, they have to like, take that ball of mud, that massive things, it's very unwieldy and they need to edit it. That's why I think that it's, it's kind of shocking to many people that, you know, some software release cycles are so slow. But so, yes, I think that, you know, there are real productivity gains. I think that it's probably not reducing the number of hours that programmers are working right now.
21:37 Speaker 03 We need to take a quick break. We'll be right back.
21:48 Speaker 01 New and exclusive holiday decor just launched online at The Home Depot. Check out the viral Grand Duchess collection with smart holiday lights you can control from your phone. Easily customize the twinkle across trees, wreaths, and garlands to match your holiday vibe. Create stunning light effects, then switch it up with the tap of your fingers. Even the man in red will be impressed. Shop the holiday decor you want, online only, at homedepot.com.
22:19 Speaker 02 This episode is brought to you by Michelob Ultra. Good times mean competitions between you and your friends, like seeing who can throw some pigskin the farthest, or rush the field the fastest. Winner gets an Ultra. Because it's not just a superior light beer, it's also giving fans superior access to their favorite sporting events. Superior is worth playing for. Visit michelobultra.com/superioraccess to enter. Enjoy responsibly. Copyright 2026 Anheuser-Busch InBev Michelob Ultra Beer, St. Louis, Missouri.
22:48 Speaker 05 Flowing ad budget on metrics that look great, till the CFO sees them? That's bull spend. And marketers are calling it out, in. Dashboard, Confessions.
22:57 Speaker 06 I remember telling my boss, "It'll be good for the brand," when leads were slow. Yeah, it wasn't.
23:05 Speaker 05 Cut the bull spend. LinkedIn lets you target by company, job title, and more. Advertise on LinkedIn. Spend $250 on your first campaign and get a $250 credit. Go to LinkedIn.com/campaign. Terms and conditions apply.
23:21 Speaker 03 We're back with AnySphere CEO Michael Truul. Before the break, we were discussing how his company's product, Chirso AI, is being adopted by professional coders. But now I wanted to ask about a different use case: vibe coding.
23:36 Speaker 03 Well, you mentioned non-technical people. Cursor is used by a lot of professional programmers, but this year saw the coining of the term "vibe coding" to describe what more amateur programmers can do, sometimes even complete novices, and often with tools like Cursor. How big is the vibe coding use case at Cursor? And what do you think is the future of vibe coding?
23:58 Speaker 04 So, our main goal is to help people who build software for a living. Right now, that means engineers. And so that's our main use case. It's been interesting to see, as you focus on that use case, and you use the understandings you get from that use case to kind of push the tech forward, and you hop programmers up more and more levels of abstraction, how it then also makes things more accessible. And that's something that we're really excited about. And I think in the end state, I do think that building software is going to be way more accessible. You're not going to have to have tons of experience on understanding programming languages and compilers. And I think there's still a bunch more work to do before anyone can build kind of professional grade software. That said, it's been really cool seeing people spin up projects and prototypes from scratch,
24:42 Speaker 04 designers in professional settings doing that. It's been really interesting to see non-technical people contribute small patches and bug fixes or small feature changes to professional software projects already. And that's kind of the vibe coding use case, not our main use case, not where the company makes most of its money, but one that I think will become bigger and bigger as you push the ceiling of focusing on professional developers.
25:04 Speaker 03 I'm curious what you think of as the demand for it, though. I understand it's not your focus of the business, and people like to talk about it. I think people, look, it feels cool to have never built software before and all of a sudden, next thing you know, you've actually created a little to-do list app for yourself or something.
25:23 Speaker 04 I probably differ from some of my colleagues on this personally. As the world as it exists right now, I do think that the two buckets of that vibe coding use case, one is there's like an entertainment bucket of you're doing these things mostly for personal enjoyment or hobbies, and then there's also a bucket that's more professional.
25:44 Speaker 04 And I think that that's like designers doing prototypes or that's people that work to serve customers, contributing back bug fixes to a professional codebase. And the way in which I probably differ from some of the people I work with is there's a group of people who are really, really, really interested in end user programming and throwaway apps and personalized software, where kind of everyone, yeah, like builds entirely builds their own tools. I think that that's really cool. I think enabling that is really cool. And I think a lot of people who aren't technical will be interested in doing that. But I still think even if you get to a world where anyone can build things on computers, there's going to be, I think, most of the use cases will still be served by like a small minority of, you know, 5% of the world that's caring a ton about the tools and building them.
26:30 Speaker 04 And that everyone will more use those tools because I just think that like the interest in that stuff really differs amongst the population. But so, yeah, right now, right now commercially, I think that a lot of the, a lot of the more vibe coding stuff falls more into like a mid journey camp or like an entertainment camp. But it's something that some people get interested in for a bit and then kind of put it aside. And then, you know, some of it is in this professional camp of people that, you know, work on software for a living, but, but don't code right now.
26:56 Speaker 03 I think you're right because when I worked at more traditional companies, whenever a new piece of software was introduced, everyone would get upset. So that's my case for most people not becoming like sort of, you know, pro, pro vibe coders. I like software though, so I'm, I'm, I'm vibe code curious. You know, maybe two or three generations from now in Cursor, I'll be able to make myself something useful. You mentioned earlier that there are these kind of two main ways that people use Cursor. There is the, I'm looking at code and you're helping me autocomplete things. And then there is the, I'm going to give you a task and walk away and come back and see what you've built. You told Ben Thompson recently that over the course of the next six months or a year,
27:40 Speaker 03 you think you can get to a place where maybe 20 or 25% of a professional software engineer's job might be the latter use case of just handing off work to the computer and having the computer do the work end to end. Any updates to that number in the last month or so? And how high do you think that number can scale ultimately?
28:00 Speaker 04 I think these things are really hard to predict. I think some of the things that are blocking you from getting to 100%, one is having the models learn new things.
28:13 Speaker 04 Like, understand an entire codebase, understand the context of an organization. But yeah, learn from their mistakes and really learn new things. I still think that the field doesn't have an amazing solution for that. The two candidate solutions are, one is you make the quote unquote context windows longer, which is these large language models, they see, they have like a fixed window of text or images that they can see. And then there's a limit to that. And outside of that, you know, it's just the, the model that came off the assembly line. And then that new kind of information that's put into the model's head, which is very different from humans, because humans are going through the world and like, you know, your brain is changing all the time, you're getting new things. That's kind of persists with you and like, obviously, you know, some memories fade away, but it persists with you somewhat.
28:58 Speaker 04 Uh, but so candidate solution number one to the continual learning problem is just make the context windows really big. Candidate solution number two is train the models. And so every time you want them to learn a new thing or new capability, you go and collect some training data on that and then you throw it into the, the model's mix. And both of those have big issues, I think. But that's, that's one thing that's stopping you. And, you know, I think that the rate of really consequential ideas in ML that are kind of like new paradigm shifts is pretty low industry wide, even though that, you know, the rate of progress has been, you know, really fast over the past five years. And so ideas of, of the form of, you know, replacing long context or in-context learning
29:43 Speaker 04 and like fine tuning with some other way of continual learning. I, I don't think that the field actually has an amazing track record of, of generating lots of ideas like that. I think it's sort of ideas on the rate of, you know, maybe one every three years. So I think that will take some time. I think the multimodal stuff will take time too. The reason that's important for programming is you want to play with the software and you want to be able to, you know, click buttons and. Actually, yeah, use the output. You want to be able to use tools also to help you make software, tools that have GUIs. So for instance, you know, observability solutions like Datadog are important for understanding how you can improve a professional piece of software. So that's, that feels like it's needed. These models also, they can work coherently for
30:24 Speaker 04 minutes at a time, now even hours in some cases, but it's a different thing to work on a task for, for the equivalent of a human's weeks. And so just even architecturally, knowing if we're going to be like coherent over sequences that long will be interesting to see. And that I think will be tricky. But there are all of these, you know, all of these technical blockers to getting to something that's 100%. And there's many more that you could list, and there are also many unknown unknowns. And I think that in a year or so, even with just playing the game of going from a high level text instruction to changes throughout a code base, playing that really well, I think if in the bulk case, you know, you could probably do over half of, of programming as it exists today. Yeah.
31:04 Speaker 03 Yeah. I see these studies that Meter puts out where they look at the average length of time that a software or that an AI model can do, and it does keep doubling at this really impressive rate. So I think the, the hurdles that you identify are super important, but when you pull back, it does seem like length of task is really improving. And ultimately, you know, humans don't tend to work on discrete tasks that are all that long. So I do think it's getting easier for people to imagine an LLM putting in a full day's work.
31:37 Speaker 04 Yeah, I think that just like forecasting these things is tricky. And one like related field that can maybe be telling of how things will evolve here is just kind of the history of self-driving, which obviously has made leaps at the bounds of advancements. And, you know, in San Francisco, we, they're, they're away most. There are commercial self-driving cars. My understanding is Tesla's also made big improvements. But, you know, I remember back in 2017, when people thought, you know, self-driving was going to be done and deployed within a year. And obviously there are still, still big barriers to getting it out into the world. And that feels like a, you know, as hard and as very distracting is, it does feel like a much lower ceiling task than some of the stuff the field's talking about right now. So we will see.
32:19 Speaker 03 Yeah, interesting. This, I do want to sort of ask you about timeline stuff, but I'm gonna, I'm gonna wait until a little bit later. All right, let me now ask you some of the famous decoder questions, Michael. How big is Enesphere today? How many employees do you have?
32:31 Speaker 04 Uh, we're roughly 150 people right now.
32:34 Speaker 03 Okay. And when you think about like how big you want the company to be, are you somebody who envisions very big workforce? Or do you sort of like the smaller, nibbler, nibbler team?
32:44 Speaker 04 Uh, we do like the nibbler team. And I think the caveat there is like, we want to keep the team nimbler for the scope of work that we're tackling, but that will still mean growing the team a lot over the next couple of years. But yeah, I wonder if it will be possible to build, you know, a thriving technology company that do, does really important work with, you know, a maximum team size of maybe 2,000 people or something like that. You know, something of the size of the, the New York Times. And, uh, you know, we're excited to see if that is, if that is possible. But definitely we need to grow a lot from, from our current headcount.
33:18 Speaker 03 What is your org chart like? You have a few co-founders. How do you all divvy up your responsibilities?
33:24 Speaker 04 The two biggest areas of the org are engineering, the research side of things, like R and D generally. And then the go-to-market side of things, so serving customers. This is a company that has really benefited from having a big set of co-founders and a, a big, very capable founding team. And so there's a lot of, uh, across that scope, kind of dividing and conquering. In particular, I think that there's a really important set of people on the founding team who have done phenomenal work in, in building out that early part of the go-to-market side of things. Uh, and a lot of that is just entirely of people in the founding team is kind of entirely credited to like a subset of it. Uh, and so there's a lot of dividing and conquering across the business.
34:08 Speaker 04 At the same time, I think actually amongst, once you like zoom into the technical side of things, there's like an intense focus from the four co-founders on that. And really putting kind of all eggs in in one basket of that side of the business. And so, you know, I think that we're lucky enough to be in a time where there are really, really useful products to build in our space. And I think that the highest, you know, the highest order bid, the thing you cannot mess up is having, having the best product in the space. And so we've been able to be relatively lean in other parts of the business, especially relative to our scale, but also as a ratio to engineering and research and still be able to grow really far.
34:47 Speaker 03 What part of the business do you like keeping for yourself? Like, where do you like getting your hands dirty and would you be mad if someone tried to take it away from you?
34:53 Speaker 04 I spend a lot of time trying to help how I can in, in growing the team. And we think hiring is incredibly important and especially the hiring of ICs.
35:02 Speaker 03 Individual contributors.
35:04 Speaker 04 Yeah, yeah, individual contributors. I think that one way technology companies die is that the best ICs start to feel disengaged, like they don't have control over the company and the talent density lowers. And then I think that if you're working on technology, like no matter how good kind of the management layer is, if just you don't, you have less than excellent people doing the real work, I think there's only, only so much you can do. Like, I think that the dynamic range of what management can do is, is kind of limited. And so I like to help how I can by spending a bunch of time on hiring. And we actually got to maybe 75 people just with the co-founders hiring and without hiring functional recruiters. And so now, you know, have fantastic people helping us hiring, people on the recruiting
35:49 Speaker 04 side of things that work with us closely. Spend a bunch of time on that and then try to help how I can on the engineering and product side of things. And those are the two biggest areas of focus. And then there's a long list of long tail things and yeah.
36:04 Speaker 03 Right. You're fairly young. I think you're 25 and have had to make a lot of really big decisions about raising money, making acquisitions, all those hiring decisions that you just made. How do you try to make decisions? Do you have a framework that you use or is everything ad hoc?
36:20 Speaker 04 Yeah, I'm not sure there's one framework. I think that, you know, some pretty common devices that help us is we try to do our best to farm from dissent, kind of all up and down the group, the org. And this is not just for me, it's try to do it for kind of all decisions in the company is having increasingly like a very clear DRI and then, you know, lots of people who are kind of inputs to the decision. Every decision is pretty unique. I think that, you know, other devices that are well-known, that are helpful are kind of understanding how high stakes the decision is and how reversible it is. I think that especially when, you know, you're in a vertical like ours with the speed that
37:03 Speaker 04 it's moving, there's just a limit on the amount of time and the amount of information you can gather on each thing. And then, you know, other devices like, you know, clearly communicating the decision and using that as a way to kind of force clarity for how you thought it through.
37:18 Speaker 03 We need to take another quick break. We'll be right back.
37:29 Speaker 01 New and exclusive holiday decor just launched online at The Home Depot. Check out the viral Grand Duchess collection with smart holiday lights you can control from your phone. Easily customize the twinkle across trees, wreaths, and garlands to match your holiday vibe. Create stunning light effects, then switch it up with the tap of your fingers. Even the man in red will be impressed. Shop the holiday decor you want, online only, at homedepot.com.
38:01 Speaker 02 This episode is brought to you by Michelob Ultra. Good times mean competitions between you and your friends, like seeing who can throw some pigskin the farthest or rush the field the fastest. Winner gets an Ultra. Because it's not just a superior-like beer, it's also giving fans superior access to their favorite sporting events. Superior is worth playing for. Visit michelobultra.com/superioraccess to enter. Enjoy responsibly. Copyright 2026 Anheuser-Busch InBev Michelob Ultra Beer, St. Louis, Missouri.
38:29 Speaker 05 Are your ad campaigns lighting up the dashboard? But not the pipeline? That's bull spend. And marketers are calling it out in. Dashboard Confessions.
38:38 Speaker 07 My boss asked for results, so I opened my dashboard for the only positive sounding metric I had. Impressions.
38:46 Speaker 05 Cut the bull spend. See revenue, not just reach. LinkedIn delivers the highest return on ad spend of major ad networks. Advertise on LinkedIn. Spend $250 on your first campaign and get a $250 credit. Go to LinkedIn.com/campaign. Terms and conditions apply.
39:03 Speaker 03 We're back with AnySphere CEO Michael Truell. We just covered the core Decoder questions, but now I wanted to talk a bit more about hiring, in particular, how the AI talent wars are affecting a company like AnySphere.
39:18 Speaker 03 Well, let's talk a little bit more about hiring, since you brought it up. There has been talk that OpenAI had considered acquiring you. And I have to ask, given his recent spending spree, has Mark Zuckerberg invited you to his house in Tahoe?
39:32 Speaker 04 No, no.
39:34 Speaker 03 No? He's not coming around with his $200 million signing bonuses saying, "Michael, why don't you kind of come over here? We're building superintelligence."
39:41 Speaker 04 This for us is kind of life's work territory. So yeah, feel really lucky to have the technology lineup, kind of the initial founding team lineup, the people that have decided to join us, and the way things have gone on the product, to have the pieces in place to execute on this ambitious goal of automating programming. And time will tell if we're going to be the ones to do that. But as people who have been programming for a long time and working on AI for almost as much, being able to reinvent programming and help people build whatever they want to build on computers with AI kind of feels perfect for us. And it feels like one of the best commercial applications of this technology, too. So I think that if you can succeed in that, you can also push the field for it in big ways for other verticals and other industries, too. And so, no.
40:23 Speaker 03 Yeah, it sounds like you really want to stay independent.
40:26 Speaker 04 Yep.
40:27 Speaker 03 Has Meta's recent hiring spree made it noticeably harder for you to recruit lately?
40:35 Speaker 04 No, I not really. The research team, we try to keep fairly small. I mean, the whole company is kind of small relative to what it's doing, but the research team especially. People think through hiring decisions in different ways. And what we have to offer is most appealing to people who want to be a part of an especially small team, working on something focused, kind of solving problems with AI out in the real world, want to be in a place too that marries. Like, we're kind of this weird company. You talked about some products that are being made by some of the great folks that work on the API models. But I think that we're like this weird experiential company that's smack dab in between the foundation model labs and normal software companies, where we try to be really excellent at both
41:20 Speaker 04 the product side of things and the model side of things, under one roof, and have those feed into each other. And so we appeal to, I think, a certain type of ML researcher or ML engineer. And for them, I think it's kind of about being part of this and a little bit less like some of the other things.
41:38 Speaker 03 One last hiring question. It was reported this week that two folks who used to run Claude Code that you'd recruited to come over to Cursor left back after a couple of weeks. Can you speak at all to what happened there?
41:52 Speaker 04 Kat and Boris are awesome, and I think that they have a lot left to do on Claude Code. And they're really, as I understand it, just the people behind that, and that is their creation. And as someone who's been working on something for three and a half years from inception, kind of understand the ownership that comes with that. And they have a lot left to do, and they were excited about that, and so they've decided to stay.
42:17 Speaker 03 Cool. It seems like you were mentioning this interesting position that you sit in, in between the big labs and other startup companies who are using your software. How do you describe Cursor's culture when you're recruiting people?
42:30 Speaker 04 Perhaps unsurprisingly, we are process skeptical and hierarchy skeptical. And so we need to, as we do more and more ambitious things, like more and more coordination is required. But for a certain level of thing, the scope of the company, we try to be pretty light on each of those. I think it's a very intellectually honest group. It feels very low stakes to criticize things and just be open, very publicly about feedback on work. It's a very intellectually curious group. I think that people are interested in doing this work for the end goal of automating programming. And separate from any work-life balance things, because we want this to be a place that's all levels of work-life balance can do great work. It's a place where I think no one really treats it, so far, like just a job.
43:15 Speaker 04 They're really, really excited about this. And I think it's a special time to be building technology. And so one should try to seek out a role where you don't treat it like just a job. I think it's very focused and understated. I think from the outside, partially because of how little communication we do with the outside world, and we need to get much better at that. I think mostly people know Cursor as, oh, that thing that grew really fast, and kind of know about top-level metrics and things like that for just how fast the adoption has been. And internally, we've thought that it's really important to hire people who are, while they might be very ambitious, are very humble and pretty understated and pretty focused and level-headed. Because there's noise left and right.
43:58 Speaker 04 And I think that, yeah, just having clear focus and putting your head down is actually really, really important for people being happy in this space, and also just for the execution of the team. Yeah, those are some things to describe the current group.
44:09 Speaker 03 You mentioned communicating with the outside world. I think Cursor's history is mostly just a history of delighting its customers. But you did have this moment recently where you changed the way you priced things, and folks got pretty mad. And basically, you just moved from a set fee to more usage-based pricing, and some people ran over their limits without realizing it. What did you learn from that experience?
44:36 Speaker 04 Yeah, I think that there was a lot to learn from that and a lot on our end that we need to improve on. To set the stage, so the way Cursor pricing has worked even back when Cursor first started is by and large, you sign up for a subscription and then you get an allotment of a certain number of times you can use the AI over the course of your subscription term. And the pricing evolves, features were added, features were changed, kind of like up and down that limit has, or like, you know, there are different ways, like you have been able to pay down that limit or not pay down that limit over time. And what's happened in parallel is kind of using the AI once, what that means, the value
45:17 Speaker 04 that gives people and the underlying costs in some cases has changed a lot. One big switch there for us is that increasingly, when you quote unquote "use the AI," the AI is working for longer and longer and longer. And so you called out that chart that you've seen where it's showing the kind of max time an AI can work, and it's gone from seconds to minutes to hours at this point, and it's gone up very fast. We're kind of front lines of that, where now when you ask the AI to go do something or answer a question, it can work for a very, very, very long time. And that changes the value it can give to you. You can go from just asking a simple programming question to having it write 300 lines of code for you, and that also changes the underlying costs, and in particular, less the median and more the variance of those costs. So yeah, we bundled together
46:00 Speaker 04 a series of pricing changes, and the one that garnered the most attention was switching from a world where the monthly allotment is in requests to it's in the underlying compute that you're spending. And one thing to knit on what you said is that actually usage-based had been a big component of Cursor before, because over the life of Cursor, people have just used the AI more and more and more and more. Then they started running out of limits, and we wanted to give people a way to burst past that. What this did is it changed the structure of also how that usage pricing worked, where it's not on a request basis, it's on the underlying compute basis. And definitely, that could have been communicated legions better. There's a lot we learned from that experience and a lot we need to show up on in the future.
46:45 Speaker 03 Yeah, I think it's hard for consumers in particular to understand usage-based pricing because they're used to Spotify and Netflix where they pay their 10 or 20 bucks a month, and it's sort of all you can eat. But the economics of AI just don't really work that way.
47:02 Speaker 04 It will be interesting to see how things play out in our space in particular. So I think that for like the consumer chat app market, so far at least there's been.
47:12 Speaker 04 Yeah, I, it would be interesting to see how the, the curves of just like how compute per user over time has gone up. But I wouldn't be that surprised if it's been pretty flat over the past, you know, 18 months or so. Where the original GPT four, you know, don't, I'm not privy to any inside information. But it seems like there have been big gains from a model size perspective where you can actually miniaturize models and get the same level of intelligence. And so I think that the model that most professional users are using in something like a chat GPT has actually maybe gotten smaller over time, the compute usage has just gone down. But in our space, yeah, I think that there's just for one user, I think the compute is probably going to go up. And there's a world in which the token costs don't go down fast enough.
47:52 Speaker 04 And it starts to become a little bit more like AWS costs and a little bit less like Percy productivity software. And still remains to be seen. But one thing to note is that we do think it's really, really, really important to offer users choice. And so we want to be the best way to code with AI. If you just want to turn on all the dials and just get the best, most expensive experience. We also want to be the best way to code with AI. If you want to just pay for a predictable subscription and get the best thing that that, you know, price can offer you. Even for the, you know, the main individual plan, the $20 pro plan. The vast majority of those users don't hit their monthly limits. And so don't get, aren't hit with a, you know, a message saying you need to turn on usage pricing or not.
48:30 Speaker 03 That's the kind of AI user I am. I never hit my, it makes me feel like I need to be using it more, you know? I, I'm trying to.
48:36 Speaker 04 There is a really, really big difference between the top 5% and, you know, a median, a median user. So some people are very, very, very AI forward.
48:47 Speaker 03 Well, coming to my last couple of questions here, I want to try to get at how AGI pilled you are. Because when we were talking earlier, you're sort of identifying all these very real technical problems in building more advanced systems that are just truly unsolved problems in AI. You know, the, the size of the context when you're giving these systems longer memory, helping them learn the way that a human. Might be able to learn. We don't know how to do that yet. And yet there are lots of folks in the industry who believe that by 2027, 2028, the world looks very, very different. So where do you sort of plot yourself on the spectrum of people who think that
49:28 Speaker 03 everything is absolutely about to change and we're sort of at the start of a process that's going to take decades?
49:34 Speaker 04 I think we're kind of this bet on the messy middle, where we do think it's going to take decades. We do think that nonetheless, AI is going to be this transformational technological shift for the world, bigger than, you know, maybe, yeah, just, you know, a very, very, very big technological shift. And when we started working on Cursor, it was funny, we'd get these kind of two dual responses. And I think one is now increasingly falling out of favor, just, you know, with the rise of the first AI products that have really reached billions of people. But early 22, we would get kind of two reactions. One reaction was, why are you working on AI? You know, I'm not sure that there's really much to do there.
50:17 Speaker 04 The other reaction that we would get, because we did have close friends and colleagues who are very interested in AI is, why are you working on, you know, insert X application, whether it be CAD or whether it be programming specifically? You know, AGI is going to wipe all of this stuff out in, you know, Y years, you know, maybe it's 2024, 2025. We think it's this middle road of this jagged peak where if you actually peek under the hood at what's driven AI progress so far. Again, I think that there's been a few ideas that have really worked. There's been lots of details to fill in between, but there have been a few really, really important ideas. I think that despite the number of people that have worked on deep learning over the past decade and a half, the rate of idea generation in the field, like really, really consequential idea generation in the field, hasn't budged that much.
51:01 Speaker 04 And I think that there are lots of real technical problems that we need to grapple with. And so I think that there's like this urge to anthropomorphize these models. And see them be amazing and human level or superhuman at some things and then think that they will just be great at everything. And I really think it's this very jagged peak. And so I think it's going to take decades. I think it's going to be progressive. I think that one of our most ambitious hopes with Cursor is if we are to succeed in automating programming and building an amazing product here, that makes it so you can build things on computers just with the minimal intent necessary. Maybe the success of that and the techniques that we need to figure out in doing that can also be helpful for pushing
51:42 Speaker 04 AI progress forward in general. And I think that the experiment to play back here is if you were in 2000 or 1999 and you wanted to push forward AI, one of the best things you could do is work on something that looks like Google and make that successful and make that R and D available to the world. And so, you know, in some ways, one of the ways, at least I think about what we're doing is trying to do that.
52:04 Speaker 03 Okay, so it sounds like you don't think that there's just going to be one big new training run with like a lot more parameters and we're going to wake up to a machine god.
52:13 Speaker 04 You know, time will tell. My best guess, yeah, and I think it's important to have healthy skepticism about how much you can know with these things. But my best guess is that it will take longer than that, yet also still be this this big transformational thing.
52:28 Speaker 03 All right. Well, last question here. We've talked a couple of times today about how hard predictions are in general. So I'm not going to ask you to do something crazy like predict what Cursor is going to look like five years from now. But when you think about it, maybe two years from now, what do you hope it's doing that it isn't quite doing yet?
52:46 Speaker 04 I think a bunch of things. So I think in the short term, we're excited about a world where you can delegate more and more work to kind of very fast, helpful, helpful humans. And you can build a really amazing experience for making that work delightful and orchestrating work amongst amongst these agents. Another idea that we've been, or I've been interested in for a long time, which is a bit risky, is, you know, I think that if you can get to a world where you're delegating more and more work to the AI, you'll start to run into an issue, which is, do you look at the code? And are you reading everything line by line? Or are you just kind of ignoring the code wholesale? And I think that neither
53:29 Speaker 04 closing your eyes and ignoring the code entirely in a professional setting, or reading everything line by line will really work. And so I think you'll need this middle ground. And I think that that could look like the evolution of programming languages to be higher level and to be less formal. And, you know, all that a programming language really is, is it's a UI for you as a programmer to specify, you know, exactly what you want the computer to do. And it's also a way for you to look at and read exactly how the software works right now. And yeah, I think that there's a world where programming languages will evolve to be much higher level, more compressed instead of millions of lines, you know, hundreds of thousands of lines of code. And, you know, I think that for a while, an important way you build software is you
54:14 Speaker 04 could, you know, read and point at and edit that kind of higher level programming language. And I think that this also kind of gets at a bigger idea that's behind the company of, you know, there's all this work to do on the model side of things. The field's going to do some of that. We're going to try to do some of that. But then the end state of what we want to do is also this UI problem of how do we get the stuff that's in your head onto the screen? And I think that the vision of you just entirely build software by typing into a chat box is powerful. Like, I think that that's a really simple UI. You can get very far with that. But I don't think it can be the end state. You need, you need more control when you're building professional software. And so you need to be able to kind of point at, you know, different elements on the screen and be able to, you know, dive into the tiniest detail and change a few pixels.
54:58 Speaker 04 You also need to be able to point at parts of the logic and understand exactly how the software works and be able to edit something very, very fine-grained. That requires rethinking new UIs for these things. And the UI for that right now is programming languages. And so I think that they're, they're going to evolve.
55:13 Speaker 03 All right. Well, a lot of fascinating things that you're working on. Michael, thank you for coming on Decoder.
55:19 Speaker 04 Thank you for having me.
55:24 Speaker 03 Thank you to Michael for taking the time to speak with me. And thank you for tuning in. I hope you liked it. If you'd like to let us know what you thought about this show or what else you'd like us to cover, drop us a line. You can email theteam@[email protected] . They really do read every email. Or you can hit me up directly on threads or Blue Sky. I'm at Crumbler on threads. And I'm CaseyNewton.BeastGuy.social. Not very catchy, is it? Decoder also has a TikTok and an Instagram. You can check those out at DecoderPod. They're a lot of fun. And if you like Decoder, please share it with your friends and subscribe wherever you get your podcasts. Decoder is a production of The Verge and is part of the Vox Media Podcast Network. Decoder is produced by Kate Cox and Nick Statt.
56:08 Speaker 03 The show is edited by Ursa Wright. The Decoder music is by Breakmaster Cylinder. See you next time.
56:17 Speaker 01 New and exclusive holiday decor just launched online at The Home Depot. Check out the viral Grand Duchess collection with smart holiday lights you can control from your phone. Easily customize the twinkle across trees, wreaths, and garlands to match your holiday vibe. Create stunning light effects, then switch it up with the tap of your fingers. Even the man in red will be impressed. Shop the holiday decor you want, online only, at HomeDepot.com.
56:47 Speaker 08 Mr. Monopoly here. Miss me? Monopoly is back at McDonald's with millions of prizes. A 2027 Jeep Cherokee, a Carnival Cruise vacation, gift cards from DoorDash, and Best Buy. Feel the play in the McDonald's app. Get your bag.
57:05 Speaker 09 No purchase necessary. 50 U.S. and D.C. age 18-plus. See PlayatMcD.com for rules and PlayatMcD.com/amoe for info on playing without purchase by mail by November 2nd. Chance card game ends November 8th. Game ends November 22nd, 2026. See rules for prize details. Monopoly is a registered trademark of Hasbro.
57:18 Speaker 10 Looking for a simple way to thank your clients or recognize employees for a job well done? A Starbucks card is more than a gift. It's a pick-me-up, a break in their day, and a reminder that you appreciate them. Whether you're shopping for digital or physical cards in bulk, Starbucks cards are the perfect gift to brighten anyone's day. Share the joy of coffee and connection when you give the gift of Starbucks. Shop Starbucks cards in bulk now at StarbucksCardB2B.com.
Previous Why AI researchers are getting paid like NBA All-Stars Next GitHub's CEO says AI coding is ‘here to stay’ Page updated October 26, 2026 · transcript and summary are machine-generated
