Opening Keynote, Michael Truell — Compile 26

[00:00:00] looking at the group of people that are assembled here today and the community we have in front of [00:00:05] us it makes me think back to when we were just getting started as a company and so i don't know [00:00:12] how many of you all know how cursor got started and what the first version of cursor looked like [00:00:18] and and when it was released but we as an ai company come from the prehistoric times [00:00:24] of january 2022 uh that's when we started working on all this stuff and the first version of cursor [00:00:29] you know wasn't that long ago but again prehistoric from ai times it was released in the beginning of [00:00:34] 2023 and the initial phase of the company's life sort of had a little bit of a detour where we were [00:00:41] really excited about working on tools for developers and working on ai coding but we we didn't do it at [00:00:46] first and the reason we didn't do it is the group of people that started this company it was just four [00:00:52] programmers four really young programmers that wanted to build things that were useful but saw the set of [00:00:59] people working on ai coding and basically were a bit nervous that we couldn't we couldn't really [00:01:04] compete at the time you know the sets of companies working on ai coding uh were already really large [00:01:10] there were dozens of startups there were big tech companies of course there were the gargantuan you [00:01:15] know sort of even though they were small at the time and are much bigger now but uh very scary and [00:01:18] intimidating ai labs and we looked at all of the people working on ai coding in 2022 and we thought [00:01:25] there's just there's no room there's not much to do people have got that covered and we kind of went [00:01:31] about our way during 2022 worked on a series of projects but we kind of caught the bug we couldn't [00:01:36] stay away from it couldn't help ourselves and we started working on cursor at the end of the year [00:01:41] fundamentally because we wanted to build a tool that we really liked and a tool that we would want to [00:01:47] use and nothing out in the market really checked that box for us and so we did what any group of [00:01:55] introverted not particularly adept at company building enterprising developers would do and [00:01:59] we kind of went in a cave and we coated in our underwear for like two weeks um and we built a [00:02:04] prototype and that was how we got started and the first initial humble goal was to build a product that [00:02:11] could be our development environment um that we could stand using like it didn't make us 2x less [00:02:17] productive than whatever we were using before we got to that goal after a couple of weeks uh by cobbling [00:02:22] together you know actually a product that was entirely built from scratch and wasn't based on the same [00:02:26] foundations that the first version of cursor later became based on we built kind of the pieces of an [00:02:31] open source code editor from scratch we built language server integrations remote ssh at the time a copilot [00:02:38] integration too because there wasn't even autocomplete within the product and we got to something that [00:02:42] we could stand using and then we sent out a beta invite list it was attached to this kind of movie [00:02:48] magic demo uh that one of one of the co-founders made and got a bunch of people signing up and hand [00:02:54] onboarded the first 20 users walked them through how to use the product got them set up sat by them [00:03:00] and the first user didn't really like the product that much this the second user didn't like the [00:03:05] product that much they actually ghost us they wouldn't talk to us anymore um and the third and [00:03:09] the fourth it was also a little bit of a tepid reaction but we soon got to a place we got a couple [00:03:13] of developers using the product and actually daily driving it every day and that started the kind of [00:03:18] fundamental loop that all through all this craziness we followed over the past few years where every [00:03:24] single day we just think about how can we make the product better for us and how can we make it better [00:03:30] for the community that we're building with and that initial set of uh 20 beta testers still email with [00:03:35] them back and forth uh the next kind of phase of the community was a discord server that we had we [00:03:40] actually hired many members that discord server and now events like this are kind of the next extension [00:03:45] of that community when i look around at what we're building now i think yeah one of the things that i'm [00:03:51] most excited about is that through all of this craziness there's a group of people whose fundamental kind of [00:03:57] core ethos is it's group of developers building for developers and it's a very deeply product focused [00:04:01] company and so we have some big announcements to go through today three big announcements before we go [00:04:09] into those announcements i'm going to talk a little bit about some of the ways in which cursor has changed [00:04:16] over the past three months and change we have a lot you know that first prototype there's not a single [00:04:21] scrap or line of code that exists within our code base from that first prototype there's also not a [00:04:25] pixel on the screen that exists at all and that happened relatively quickly because code wasn't [00:04:29] very good because on the screen weren't very good but there's been a turnover time and time again in [00:04:34] cursor and you know our goal is really to build something that we'd find useful that uh developers [00:04:39] in the community find useful and with the pace of change of the technology that's changed many times [00:04:42] over and that's definitely changed recently the first big theme that we thought over the past couple [00:04:47] of months and ways in which we wanted cursor change is we wanted cursor to become simpler and more powerful [00:04:53] and in particular really to be fundamentally built for working with agents to be agent first and that's [00:05:01] a transition that's happened over the past year and a half but it's really i think finally come into [00:05:06] place over the course of the last three months and so for instance at this point over 95 of users of [00:05:12] cursor like use us primarily as an agent the vast majority of our business is being a coding agent at this point [00:05:19] uh even on a request basis agents are used about 5x more than any assistive features for instance like tab [00:05:26] and then when you control for lines of code it's many orders of magnitude more usage on the agent side [00:05:31] of things to get there there's been lots of changes both on the product side of things on building really [00:05:36] great harness side of things on the model side of things and in particular over the course of the [00:05:41] past month or two the way in which people experience cursor the pixels on the screen has changed a bunch with the release of cursor 3. [00:05:45] this is what cursor looks like now and fundamentally it's agent first you have agents in the left you have agents in the center [00:05:53] it combines you know the power of a development environment that's fundamentally built for working [00:05:59] with agents with also all of the power you expect in a professional development environment [00:06:04] and so you still have the ability to edit files you still have great remote ssh support you have language [00:06:09] servers running always and doing the normal linting operations you would expect from a professional [00:06:12] development environment so there's no compromise there and in addition i think over the past couple [00:06:17] of months we've built a really great set of pro features for working with agents and for instance that [00:06:24] includes things like design mode where you can gesture at the screen that you're building with the agent and [00:06:29] you can ask it to change a particular element of a website and you know this is part of a broader theme [00:06:33] but we want working with agents to feel like working with colleagues you should be able not just to [00:06:38] talk back and forth with them over text chat but point at a screen together cruiser 3 comes with [00:06:43] you know advanced features also like recursive sub-agents you can have sub-agents that call sub-agents and [00:06:47] have whole teams of agents at your disposal you can see everything in the context and many more features like this [00:06:53] a big part two of this next phase of the product has been getting to a place where agents can do entire [00:07:02] features entire bug fixes and entire projects for you and i think that most developers haven't gotten to [00:07:09] this point most people whether it's an art product or competitive tools they're working with agents and [00:07:14] handing them bits of work 30 minutes worth of work an hour worth of work and they're running three to five [00:07:19] locally on their computer and having to make local tool calls and that's obviously not the future like [00:07:24] you're not going to have coding asi and just be running all these agents locally on your computer [00:07:28] and having them bottleneck by falling over each other on the same code base we need to get to a world [00:07:33] where working with agents is really like working with a colleague and you can treat it just like [00:07:38] you have your own team of engineers and you need to be able to hand off whole projects and have the [00:07:43] agent go work on them for days and then come back to you with the thing just done completed and tested [00:07:47] and so big part of that has been investing in giving agents their own computer and making that experience [00:07:53] especially simple especially fast and especially powerful and so cloud has gotten a big upgrade and a [00:07:59] bunch of performance improvements and is a big part of cursor 3 as well another big change in cursor is going [00:08:08] from being this first party product that just is a closed-walled system that humans interact with [00:08:14] with really being a platform coding agents are fundamental infrastructure for teams they're fundamental [00:08:19] infrastructure for businesses they're also you know how everyone is coding today uh and with a tool that's so [00:08:26] central to your work and so central to a team's work you need to be able to change the tool uh and [00:08:32] you know edit it and um you know control all the aspects of it that you would like so we were excited [00:08:38] to announce our sdk we think it's really important that cursor is not just something that humans can [00:08:44] use directly by interacting with the product but something that's accessible with the best ux and formal [00:08:49] programming languages something that's accessible by other agents too hackability and building a platform [00:08:54] doesn't just come from the sdk it also comes from letting cursor use all of the tools that you use [00:09:01] as we work towards having agents work like colleagues they need to be able to use the same [00:09:06] tools that an engineer uses so they need to be able to trawl around datadog or pull from a database [00:09:10] or use the same design tools that you have access to and we have been excited to see the growing and [00:09:15] vast space of plugins that people have built for cursor both internally just ad hoc for their own tools [00:09:21] but also for very popular tools that are used across the user base and then a big component of making [00:09:25] cursor increasingly hackable and extensible is also investing more in the cli and so over the [00:09:31] past couple of months there had been over 50 individual quality of life and polish improvements [00:09:36] to our command line interface which you can use either in headless mode or use as a surface in and [00:09:41] of itself we think it's very important that cursor is available anywhere where you want it to be [00:09:45] available you know on web on a desktop app but in many cases there will always be a time when you want to [00:09:50] use a terminal ui and then lastly a big focus for us over the past few months has been building models and this is something that we've done ever since the [00:10:01] started the company um in 2023 not too long after the first packed together prototype you know where [00:10:08] 18 of the 20 beta testers kind of ran away from us kicking and screaming for some reason our ambitions [00:10:13] uh went higher from that boost of confidence we got from the user base and we uh started working on [00:10:18] models and so in 2023 that started with with tab models and the series of tab models we released in 2023 [00:10:23] and 2024 i think sneakily were actually some of the most popular coding models um over the course of those years and wrote a lot of the world's code [00:10:31] the next progression in our journey of building models was starting to make the first prototype agents [00:10:38] work in 2024 and 2025 and the way we got that to work was kind of by hacking around the api models [00:10:44] and building small surrogate models that could patch their weak points like when they couldn't do tool [00:10:48] calling when they couldn't search the right code base we built models to give them those capabilities [00:10:53] but over the past year our ambition has grown a ton and we've started working on agent encoding models models [00:10:59] models like you know are fundamentally foundation models that can do all of the actions that you [00:11:04] would want cursor to do and so that has been our composer series of models it started with composer one [00:11:10] which was released in november 2025 and with composer one we got rl working at scale on the workloads [00:11:19] that we care about and we started with an open source base maybe famously or infamously um and [00:11:25] uh and did did lots of work and dump in lots of compute uh to make it more useful for developers we then [00:11:32] scaled that up that same process up with composer 1.5 with composer 2 there were kind of two big paradigm [00:11:38] shifts one was getting continued pre-training working uh and not just doing the rl side of things but [00:11:44] training on tens of trillions of tokens on top of the open source models and then the other was getting [00:11:49] real time real time are all working so having a new model that's updating every five hours if we learn [00:11:53] online from what people like what people don't like where the models are being productive where [00:11:57] they're not being productive and then we scale that up with composer 2.5 and we have been really [00:12:02] hardened to see the fast interest that composers got in and that's come from a couple of things [00:12:08] and so far i think it's come from a focus on speed and a focus on cost and this is really really [00:12:13] important to us we think it would be very sad if the models got very capable if it was possible for [00:12:19] anyone to build anything they'd like on computers with a giant asterisk on it where those abilities [00:12:24] were just gate kept to the largest companies that had lots of capital to spend on these models we think [00:12:28] that that would be a sad world we want individual developers solo hackers to be able to use you know the [00:12:36] best performing models that the market has to offer and so we've focused a bunch on cost and on speed [00:12:42] so far with the latest set of composer model releases and we're seeing people start to notice [00:12:47] that it kind of makes sense in steady state to have a couple of models that you use maybe a daily [00:12:50] driver that's more price performance and then going to more expensive models sometimes when it makes [00:12:55] sense i will say our ambitions are to build just overall the most capable coding models the most useful [00:13:00] coding models overall but it's been nice to see the focus on speeding costs starting to take hold and so [00:13:06] with that we have three really big announcements about some new things that we're working on [00:13:12] and so to hop right into it i'm going to hand it over to kevin who leads one of our product teams [00:13:18] thank you michael today i want to share a little bit about how we're thinking about bringing coding [00:13:24] beyond your laptop as we've seen with fable and a bunch of these early agent systems agents will soon [00:13:31] be able to run for not only just hours but days and weeks at a time if you think about this as critical infrastructure [00:13:40] your agents need to always be on and they need to always be running we don't run production servers [00:13:46] off of our laptop for a reason what you really want is as michael mentioned a teammate that you can turn to [00:13:52] whether that's when you're out visiting a customer or on a run or an idea strikes you right before [00:13:58] you're about to go to bed a teammate that you can turn to who can actually go and implement that feature [00:14:04] request who can make it real who can deliver that a fully productionized version right when the idea [00:14:09] hits we've been working on cloud agents to deliver just that and i want to talk about a few of the design [00:14:17] principles that we think about as we've built out cloud agents the first is that cloud agents need [00:14:23] to be autonomous they need to always be on and they need to work from anywhere so we have some really [00:14:31] exciting updates across these three different dimensions that i want to talk through to make [00:14:35] agents autonomous we gave them their own dev environments so they can work just like you all do [00:14:43] they have their own computer and their own resources so they can actually test the code that they're [00:14:48] working on they can deliver demos and they can act just like you all do as you build software so what [00:14:56] goes into a dev environment well there are a few different components to this the first is that [00:15:01] obviously agents need code so they need cloned repos they need installed dependencies they need credentials [00:15:07] to be able to access build systems and internal tool chains and on the other side of that you [00:15:13] unlock this new set of capabilities for what agents can do for you they can run those tests they can [00:15:19] verify their work they can produce screenshots and things that you can visually interact with and [00:15:25] circle and describe and continue to prompt and so it is far more interactive than the agents that you [00:15:31] all may be used to working with behind me you can see what an autonomous agent working in the cloud can do [00:15:39] it's produced a demo of the work that it's just done you can review this you can circle it you can interact [00:15:46] with it it makes it so much more real when an agent has their own developer environment second [00:15:53] agents need to always be on instead of prompting an agent to do work we're moving towards an era of [00:15:59] loops and agent systems where agents are going to continue to be working on our behalf at all times they'll be [00:16:07] kicked off programmatically and they may even prompt us more than we prompt them this is only possible if [00:16:14] you have really fast and really reliable agents building always on agents is a hard infrastructure [00:16:20] problem and we are learning that the hard way over the last few months we've made cloud agents three [00:16:26] times faster and we've hit three nines of reliability our goal is ultimately to provide you all with that [00:16:33] teammate that reliable cloud agent that can do all of these new amazing things equipped with its dev [00:16:39] environment these reliability and performance gains really show up when your agents are working 24 7 for [00:16:46] you we built automations and sdk to be able to make these loops really easy to build and deploy and to [00:16:55] be able to run in the background one of my favorite automations comes from a customer amplitude who is [00:17:01] currently in the process of migrating 20 000 different instances of react components and [00:17:08] swapping them out for their tailwind equivalent this is all running in the background via an automation [00:17:14] and a custom migration agent that they've built and deployed on top of cursor and it's not just [00:17:20] amplitude we've had over 6 million automation runs since releasing our automations product just a few months [00:17:27] ago and finally you should be able to work with your agents anywhere you can now kick off agents from the [00:17:34] apps that you're most familiar with whether that be working with a colleague in slack or microsoft teams [00:17:41] whether that be a new ticket that comes into jira or linear also have a cloud api and web hooks to be able to [00:17:47] invoke these cloud agents and of course you can then work with those agents across these different surfaces [00:17:53] as michael showed from our new agents window to your editor and ide to the cli in the terminal but today [00:18:02] we're really excited to introduce cursor mobile in beta this has changed the way that we have been [00:18:09] building cursor with cursor with cursor mobile you can now work with agents on the go to unblock them to [00:18:18] dream up new ideas and to build freely from wherever you are you can see all your agents running you can [00:18:26] see which ones are blocked you can connect to all of your repos you'll also be able to see the artifacts [00:18:32] that the agent is producing gives you short verifiable screenshots for you to be able to interact with [00:18:38] on the go if anything looks off you can annotate that screenshot you can prompt your agents and you [00:18:44] can get them back to work and for those of you who want to continue to work locally maybe your desk looks [00:18:50] like a bunch of mac minis right now or you really just love juggling your laptop half open cursor mobile [00:18:56] will include remote control to be able to work with your local agents while you're on the go we're [00:19:02] making cursor for ios available today to anybody in the audience so i see everybody pulling out their [00:19:09] phones already please scan the qr code behind me to get access to the test flight and we're really [00:19:15] excited to be opening this up to the world in the coming days we cannot wait to see not only what you [00:19:21] build with cursor mobile but also where you start to build from next i'm really excited to hand it over [00:19:28] to tamas to talk about another surface that we're building thank you kevin and hi everyone i am so [00:19:38] excited to be here today so my name is tamas i'm one of the co-founders of graphite and i joined cursor [00:19:43] through our own acquisition this past january for those of you who have built software on teams you know [00:19:48] that building software is about way more than just writing code it's about testing reviewing merging [00:19:54] and deploying that code at graphite we built best in class tooling to accelerate those parts of the [00:19:58] workflow we serve the best and most demanding engineering teams including shopify snowflake [00:20:03] notion figma and countless others and they depended on us every day to power their developers in these [00:20:10] critical workflows and keep everyone productive however over the past few years we've noticed a trend [00:20:15] as these companies adopted ai tooling the tools that they relied on started to become unreliable that's [00:20:21] because over the past few years ai tooling has totally changed our industry it's enabled every [00:20:26] developer to be a 10 to 100x developer but that change has required fundamentally different tooling [00:20:32] that's why when we were acquired by cursor we accelerated our most ambitious project to rebuild that [00:20:38] tooling from scratch today i'm incredibly excited to announce origin our agent native git platform with origin you [00:20:53] thank you with origin you and your agents can create repos share code and manage changes all within the same [00:21:00] ecosystem that you already use and love it's scalable extensible and built to keep code moving so let's [00:21:08] talk around what agent native really means when we sat down and started to think around that the first word [00:21:14] that came to mind was scale as we've seen developers adopt this ai tooling we've just seen so many more [00:21:20] lines of code commits pull requests and so we went back to basics and architected a novel gate architecture [00:21:28] we leveraged the primitives offered by the cloud providers to provide better scalability reliability [00:21:33] and performance so that we can keep up with these changing demands in early load tests we've simulated [00:21:38] thousands of agents and what we've seen is that we can push and pull at the same time to a single repo [00:21:43] enough to keep up with current demands and with what we expect future demands to be the results are [00:21:47] incredible and give us confidence that origin is a rock solid piece of git infrastructure built for scale [00:21:53] both human scale and agent scale the second thing that came to mind was extensible as michael just [00:21:59] told us cursor is a platform and so as we started to think around what it would mean to build an agent [00:22:04] native git platform our first thought was that we had to make it so that you could build whatever you [00:22:10] wanted on top of it through our apis our mcp and third-party app platform you and your agents can use [00:22:16] origin for whatever works for you with generous rate limits and a comprehensive api you are in full [00:22:21] control of your own data the third thing to know about origin is that we built it to keep code moving [00:22:27] powered by the same intelligence that powers cursor origin can resolve merge conflicts fix ci failures and [00:22:34] address comments it automatically figures out next steps for each pr and only tags you in when it needs to [00:22:42] by leveraging the agents to fix and review prs it can keep code in motion and more than half time to [00:22:47] review origin is already live on cursor.com both internally and for select design partners and it [00:22:53] will be rolling out to everyone this fall if you're interested in building the future of collaborative [00:22:58] software development with us sign up for the waitlist at cursor.com origin thank you and back to michael [00:23:06] thank you tomas and one final announcement we're excited to talk today not just about the product [00:23:13] side of things but about our next model and we are very happy to announce that we are late in the stages [00:23:20] of training a very different type of model than we've ever trained before and it's different in three [00:23:25] ways first it's big and it is as big as opus and gpt second it's trained from scratch so it's not [00:23:33] starting from open source basis we love open source and we want to find more ways to contribute to open [00:23:38] source but starting from scratch will let us control all of the behaviors of the model and tune it even [00:23:43] more for the workloads that we and you care about crucially also we are running on 10 to 20x more [00:23:51] compute than we've ever had access to our last model used a little bit more compute than we had had in [00:23:56] the past but this is really the first model where we're scaling up compute a ton and this is a very very very [00:24:02] very big deal because in the past our composer models from composer 1 to composer 2.5 they were trained [00:24:09] on a very small set of gpus compared to frontier labs and that really limits what you can do the [00:24:15] things that we were doing to make the model better and better and better were fundamentally blocked by [00:24:19] whether we could run on more gpus for more hours and so this scale up of 10 to 20x uh really lets us we [00:24:26] think get to frontier and hopefully soon leapfrog it and give you all exciting new powerful capabilities [00:24:33] and then last this is going to be a model that's going to be intelligent beyond just coding we think [00:24:39] that this is important for what cursor would like to do we want to make it so that anyone can build [00:24:44] anything they'd like on a computer and the bottleneck to that is starting to become not just the code [00:24:49] writing side of things but everything that you would want out of a an engineer colleague and that means [00:24:54] using tools that engineers would use that means long-range planning that means actually testing the [00:24:59] software and clicking through buttons that means also having great ux around showing users exactly what [00:25:06] was changed by the agent but we're going to take a step outside of just coding and make this model [00:25:12] much more generally intelligent we're deep into training it's been kicked off and will be released in the next [00:25:16] couple of weeks and all of this is backed by our our partnership with face x which as you all know is [00:25:22] a little bit more than a partnership um it's been a slow news day um i think that this is both exciting [00:25:31] for this particular project uh and kind of the next series of model releases that we're looking forward to [00:25:37] but also kind of hearkening back to what we were talking about at the start there's so few players and [00:25:42] institutions in ai that can really co-design both the product side of things and the model side of [00:25:48] things and i think the fundamental company dna of the other players are all that of you know maybe [00:25:53] large tech companies or maybe things that started out as labs and then kind of backed into the developer [00:25:59] and builder side of things and this is a company that you know for better for worse i think for better [00:26:05] is really fundamentally about developers that's what we started with that's what we care about this [00:26:11] company is made up of fundamentally people that want to build a tool that's useful for them and so this [00:26:17] next phase of the company i think elevates for the first time a group of people with that product ethos [00:26:23] with that developer ethos down to its core and gives that group the ability not just to edit the pixels [00:26:29] on the screen but also to edit the fundamental capabilities of the models which is an important [00:26:34] piece of the product and so really excited about both this next release and just the series of [00:26:38] releases we're going to see over the next few months and so there you have it three big announcements uh [00:26:42] the app mobile app being released today the start of taking on github and a first sneak peek [00:26:48] the first time we've ever talked about it of this uh this new model and our plan for models going forward thanks [00:26:59] Thank you.