Machine-generated source transcript; recognition errors remain. Timestamps and subtitles use the selected source recording’s clock, not clip-relative time. Section links use cited source moments; sections without a verified moment return to the talk start.
0:19:42Let me just get the clicker real quick.
0:19:48Sweet, let's see if these slides show up first.
0:19:52All right.
0:19:53Hey everyone, I'm Jason from the developer experience team at OpenAI, and I'm super excited
0:19:58to kick off this track on
0:20:00raising your ambitions. You know, I joined in March and the Codex app wasn't
0:20:04really a thing for a couple weeks. There was no Astra yet. And when these things launched,
0:20:09I kind of realized that I started becoming someone that worked on the developer experience
0:20:13team and I worked on the everything experience team now that AI can be a tool that allows
0:20:17everyone to make things. And so I want this talk to not only to be addressable to the people
0:20:22here in this room as developers, but for everyone online as well.
0:20:26You know, in this year we've done a lot of different things, right?
0:20:28We wanted a better coding app, we built the Codex app, we've been contributing to things
0:20:32like biology, folks have wanted to make personal applications, so we launched sites, cybersecurity,
0:20:37voice, computer use, and in this past couple of months, we've done some really ambitious
0:20:41things like our contributions to math and physics and a lot of new physical engineering
0:20:46capabilities.
0:20:47It's kind of crazy to think that a couple years ago we were doing grade school math,
0:20:50and now we'll be able to like tackle Navier Stokes.
0:20:54So when I got asked to talk a little bit more
0:20:56about what ambition meant,
0:20:58I think the question I keep going back to is like,
0:21:00ambition isn't sort of how you do things,
0:21:04but like how you approach these things, right?
0:21:06What do you want to make?
0:21:06And how ambitious are you in thinking
0:21:08about the things that you make?
0:21:10I know this is a coding conference after all,
0:21:12and many of the talks here will talk about things
0:21:14like how to run like longer running tasks,
0:21:17or I think talk about dark factories and multi-agents.
0:21:20But I think from someone like me,
0:21:22it might not be as relatable to just say
0:21:24that I have thousands of agents running in the background.
0:21:26I have unlimited tokens.
0:21:28And so really I want to give a talk
0:21:30that talks a little bit more about
0:21:32what I think is the role and responsibility
0:21:35for developers in the future.
0:21:38And so using more tokens isn't what I mean
0:21:40by being more ambitious.
0:21:42To me, in a world where execution is now free,
0:21:45I think ambition is summed up by these three principles, right?
0:21:49Do you actually believe that you've created this work
0:21:51and that you have ownership of this work?
0:21:53Can you describe what your tastes are
0:21:56and understand how you can tell the model to make things better?
0:21:59And can you describe what good looks like?
0:22:00And then lastly, are you actually curious enough
0:22:03to use AI to pursue other fields, right?
0:22:06Are you actually able to expand your horizon
0:22:07rather than thinking that you're going to be replaced
0:22:09or something like that?
0:22:10And this mostly comes from three different fears
0:22:14that people have when it comes to like using AI more.
0:22:17Are we just gonna create more slop in the world?
0:22:20You know, is AI gonna destroy my ability to think critically?
0:22:23And is AI gonna replace us?
0:22:25And so I wanna talk about these things one at a time,
0:22:27and use that as a way of framing this conversation about ambition.
0:22:31And so the first thing is you really wanna think about
0:22:33owning the work that you ship.
0:22:36So the first question is like, are we gonna ship more slop?
0:22:38Right? And probably the answer is yes,
0:22:41but I try to think about this the same way that we thought about
0:22:43the camera when it first came out.
0:22:46As a raise of, like, show of hands,
0:22:48like, how many people here feel like the technical work
0:22:50they do is a creative pursuit?
0:22:53Yeah. You know, me too.
0:22:57These days, I don't think of myself as a developer.
0:22:59I mostly am thinking about myself as someone
0:23:01who's trying to will things into existence, right?
0:23:03And I have to, again, feel this ownership
0:23:05and describe what I like.
0:23:07In a past life, not many people know this,
0:23:09I used to be a machine learning engineer,
0:23:11I used to be a physicist, and I used to be an artist.
0:23:13And the story of the camera is something I keep thinking about
0:23:16as we introduce more ways of doing this kind of replication,
0:23:19as this way of creating things.
0:23:22When the camera first came out,
0:23:24painters who were busy focusing on their ability to recreate images,
0:23:28yeah, they were a little stressed out.
0:23:29But at that time, people weren't even thinking about
0:23:31whether or not photography could change the art world.
0:23:34We weren't thinking about what a movie meant.
0:23:36You see a Charlie Chaplin film when it comes out.
0:23:38It's really easy to dismiss that as some, you know, silly thing.
0:23:42But at some point you realize that not only is photography art, but it has changed the
0:23:48art world, right?
0:23:49And now we have film and all these other mediums and there's amazing things that you can explore.
0:23:54Walter Benjamin says this in a book, The Work of Art in the Age of Mechanical Reproduction,
0:23:58a book that came out right when the camera became popular.
0:24:00He says, photography freed the hand of the most important functions and brought them
0:24:05to the eyes and to the voice.
0:24:08This is kind of the same with AI now,
0:24:10so I just get to use voice mode on chat
0:24:12and just say, you know, do this and do that,
0:24:14in the same way that at some point,
0:24:16if you're a photographer or director,
0:24:18you can just say, you know, move that camera over there.
0:24:20I want a wider shot.
0:24:21I want to capture some kind of feeling,
0:24:22and maybe have a team around me
0:24:24that helps me capture that vision.
0:24:28And now we live in a world where everyone has a camera.
0:24:30And sure, maybe I have a thousand photos in my background,
0:24:33in my camera roll that I never look at.
0:24:35maybe that is a slot.
0:24:37But it only matters whether or not I have captured
0:24:39a couple of really important moments in my life.
0:24:41Right? And having ownership over those moments
0:24:44is the thing that's important.
0:24:45Like for example, like 30 years ago, like,
0:24:46I didn't really have childhood photos,
0:24:48because we just grew up in a place where we couldn't afford a camera.
0:24:50Right? And now technology has allowed everyone
0:24:53the ability to have that kind of access.
0:24:54And I sure would have liked to have a camera back then,
0:24:56even if it meant a little bit more slot.
0:24:59Right? And so if you go back to this idea of this painter,
0:25:02I think the question you have to ask yourself is,
0:25:04do you identify yourself as the painter in this example
0:25:06or as the artist?
0:25:07Right?
0:25:08Maybe it means that painting photo realistically
0:25:10isn't the skill that's really valuable,
0:25:12but if you think about yourself as the artist,
0:25:15right, whether it is the paintbrush or the camera,
0:25:17whether it is code or clay or anything else,
0:25:20these are not just mediums that allow you
0:25:22to express yourself and to make things.
0:25:25Right?
0:25:26I don't make a table and I say, like,
0:25:27DeWalt made this table.
0:25:29Right?
0:25:29Just because a director has thousands of people on set,
0:25:32it's still the director's movie,
0:25:33and even in open source, we might have thousands of contributors or even agents now running the system,
0:25:38but it's the core maintainers that help hold the responsibility for the quality of the work.
0:25:44And I don't think the danger these days is whether or not AI is going to create more slop,
0:25:49but instead it's about self-erasure.
0:25:52Are you going to be a passive observer of the work that's being created, which results in the slop,
0:25:56or are you going to be someone who is an active participant in the act of creation?
0:26:00I think many people have been sort of moving themselves away from this.
0:26:04And the only difference between a person who's making 1,000 PRs versus another is whether
0:26:08or not they actually care about the work and whether they are an active participant.
0:26:13And so as an engineer, what this means is you just have to use your products more, notice
0:26:17when things break, and decide whether things need another pass.
0:26:21A lot of the great examples you see online are not one-shot examples.
0:26:25These are things that you iterate over time because you feel proud to create this kind
0:26:29of work.
0:26:30You know, internally we have a lot of amazing work with, like, our ownership over, like,
0:26:34the Codex app, right?
0:26:36Most people here have an incredible sense of ownership over the product, and as things
0:26:39slip as we ship more, we want to bring them together.
0:26:42And again, this is mostly because of the fact that we care.
0:26:46So in terms of worrying about slop, I think that the medicine for slop is really just
0:26:52the fact that you really need to believe that you own this work.
0:26:56I think the second thing to address is whether or not you feel like you can actually describe
0:27:02what good looks like. And for that, you need to go out and be inspired.
0:27:07So the second question people are asking now is, like, do you think AI is going to destroy
0:27:11our ability to think critically, right? If I own the work, how do I know if something's
0:27:15good enough? If I just hand you a camera in this analogy, like, does it mean that it takes
0:27:20away all the effort of creation? I don't think the answer is yes. Right?
0:27:24Instead, it's now about having language
0:27:27to describe what's going on.
0:27:28So when I first started art school,
0:27:29maybe I don't know how to develop film,
0:27:31I just use my iPhone,
0:27:32but I still need to understand the elements of design
0:27:35and what does beautiful look like.
0:27:38Space, value, shape, texture.
0:27:40I would just spend a lot of time thinking,
0:27:42you know what, I don't think I understand
0:27:43what negative space means.
0:27:44Let me go do a lot of sketches and explore a little bit.
0:27:47It's the same thing with development now.
0:27:49I can make multiple apps and I have to think about
0:27:51why I like something and why I don't.
0:27:54And most of my time in art school was just thinking,
0:27:56man, like, I can make things and I can look at them
0:27:58and I just say, I don't like how this looks.
0:28:01And what I was noticing was that my taste
0:28:03and my judgment exceeded my skill.
0:28:05I could not do the thing I liked.
0:28:07And I can't just look in the mirror and tell myself,
0:28:09like, make it pop a hundred times
0:28:11and know that it's gonna be better.
0:28:12I have to go out and develop that language.
0:28:15People say, like, taste is the moat,
0:28:17but I think I generally hate that line
0:28:19because it doesn't tell you what taste really means.
0:28:22To me, taste is the ability to describe what you like
0:28:24and predict what others will like.
0:28:26And building this model of the world and of other people,
0:28:30that's the new thinking.
0:28:33Winstein's, the limits of my language means the limits of my world.
0:28:36And I think that's very much the case on pursuing ambitious projects.
0:28:41You know, there's a funny example of just like the use of language.
0:28:44I don't know if Steve was in the room, but he made a tweet a couple days ago
0:28:46that just says, find all the implicit state machines in my app.
0:28:49I had to go Google what that meant, right?
0:28:52And all of a sudden, all of the comments are just how, like, everyone's apps are, like,
0:28:5420% faster.
0:28:56These are now, like, really simple ways of just developing some language to understand
0:28:59what good looks like and how to make things better.
0:29:03If you think of, like, model progress, in machine learning, there's a really helpful
0:29:06concept called the generator-discriminator gap, right?
0:29:10You'll have two models.
0:29:11One grades the model, and then one actually creates the outputs.
0:29:14And what happens is, is if the grader is too good, then the model can't learn.
0:29:18And if the generator's too good,
0:29:19the discriminator can't help the model improve.
0:29:22We are now that discriminator, right?
0:29:24As model capabilities go up, at some point,
0:29:27maybe the model's output hits your own taste levels.
0:29:30Then it's really hard to actually improve.
0:29:32And then to make things that are better,
0:29:34you have to go out, develop that language,
0:29:36and know what good looks like.
0:29:37And the failure mode is here,
0:29:39just not being able to discern
0:29:42whether or not a model can improve.
0:29:44Right?
0:29:45And again, what this means is we have to go out
0:29:47and learn again.
0:29:49You have to think again.
0:29:50This is an example of a bunch of things
0:29:52I spent my past year learning, like in just building software.
0:29:55Or like, as I build these applications,
0:29:56I realize, oh man, these websites look really ugly.
0:29:58Like, I need to learn about typography.
0:30:00Like, why is everything italic? Like, I don't understand.
0:30:03And I spent a lot of time learning about typography. Then it was about, like, how the information
0:30:06was presented to avoid looking like slop. And then it was about, like, motion design, and
0:30:11how can I make things move in a beautiful way, and I got to spend so much time learning about
0:30:15animation, and then learning about, like, interaction patterns, and what makes an app feel beautiful.
0:30:19And this is really fun, right? And this is, again, spending more time learning in order to
0:30:24make great things.
0:30:26And most importantly, having the ability to describe things is what allows you to differentiate,
0:30:30right?
0:30:31If you think about these purple apps, the reason these existed was because the people making
0:30:35these applications, like, didn't have the site.
0:30:38They couldn't actually describe what was wrong, and they thought this was okay, right?
0:30:41And this limiting in language is sort of the problem that we face.
0:30:46And then so the question now is, okay, how do we develop this new language?
0:30:49And honestly, the line that I've sort of been like punished to now is just like, if you want good taste, you have to eat.
0:30:55You have to go out and explore things, right?
0:30:58People are like vibe coding video games.
0:30:59Maybe the answer isn't just to like tell the model to do better, but like play some video games and understand what feels good, what animations feel good, what stories matter, what stories do I want to tell, right?
0:31:09If you're building tools, try other people's tools and applications and understand what you like.
0:31:14Look at beautiful things.
0:31:15I've been doing a lot of like 3D printing with AI,
0:31:18and everything just looks like rectangles glued together.
0:31:21And I just kept saying like, make it more beautiful.
0:31:22And that's a really frustrating experience.
0:31:25And then I started looking at like sports cars,
0:31:27like industrial design.
0:31:28I started going to stores and looking at things
0:31:29and going like, let me describe and understand
0:31:31what like good curves look like.
0:31:33And again, it's like a really fun experience.
0:31:35And now more of my designs are like incredibly beautiful,
0:31:38for what I think is beautiful, right?
0:31:40I've been helping with like marketing,
0:31:41and I realized that like I tell the marketing team
0:31:43that they should use the Codex app more,
0:31:45but I'm not watching more ads.
0:31:47And as I'm watching more movies,
0:31:48I realize there's great stories to tell,
0:31:50and everything becomes better again.
0:31:51And this is kind of the really exciting part
0:31:54of using these tools.
0:31:57And thinking about why I like things
0:31:59and why things are amazing,
0:32:00why things feel good now is how I spend my time.
0:32:03So I'm not thinking less, I'm thinking more,
0:32:05and it's really enriching my world
0:32:07and making me see things differently.
0:32:09And if you're running a team,
0:32:10I think the one takeaway here is that
0:32:12Like, if you really believe that AI is making you more productive,
0:32:16I think you owe it to yourself and the rest of the team
0:32:19to make space to go out and explore things
0:32:21and try new applications
0:32:22and just actually go out and, like, experience the world, right?
0:32:26And if you are someone that makes really great work,
0:32:28inspiring others as well, right?
0:32:30And that's how you can come back with a point of view.
0:32:34So go out and be inspired.
0:32:35Learn to be a better critic
0:32:36and recognize that you're not actually thinking less,
0:32:39you're thinking more.
0:32:41And the last thing is actually just being curious,
0:32:43really, really curious to pursue new ideas.
0:32:46Like one question I want you guys to think about
0:32:48as you are thinking about your role as software engineers
0:32:50is like what were the things you gave up
0:32:52or put aside to pursue software, right?
0:32:56Where else could use your judgment and your experience
0:32:59you've developed as a software engineer?
0:33:00Because I think people worry about being replaced,
0:33:02but really I think people are needed in way more places.
0:33:06If you think about when the, you know,
0:33:08invention of like the pencil or sculpting became more easy, all these different industries
0:33:13in the creative world became really very much multidisciplinary, right?
0:33:19It completely made sense that if you were a furniture designer, you could be an architect
0:33:23and you could like make a movie and design kids' toys and playgrounds, right?
0:33:28Like Michelangelo is a sculptor and a painter and an architect and a poet, you know?
0:33:32It completely made sense that you could do all these things, right?
0:33:35And this is a creative world.
0:33:36Like you invent the pencil, you can do a lot of really amazing things.
0:33:38But with technology, it's become even more complicated to do so.
0:33:42I just can't get, like, four PhDs.
0:33:44And if I could, like, maybe I'll just be, like, 50.
0:33:47Right?
0:33:47And so being technical has become incredibly difficult.
0:33:51But when you do cross-pollinate these different industries, amazing things happen.
0:33:55Right?
0:33:55The physicist was, sorry, the founder of the molecular biology was a physicist.
0:34:00Right?
0:34:01A mathematician started leading the human genome project.
0:34:04Right?
0:34:04And the mathematics around magnets is what is used to help understand neural networks.
0:34:09But this is now, like, super hard to do, because you need so much information.
0:34:14Right?
0:34:14Even in our own world, like, we have a bunch of AI helping us design, like, chips and writing software.
0:34:18And these, again, are just very multidisciplinary pursuits.
0:34:22If you look at what it means to be ambitious outside in the real world,
0:34:26there's a machine learning consultant using AI and molecular biology to, like, help treat their dog's cancer.
0:34:33There was this public servant in Hokkaido who used, like, chat and, like, computer vision
0:34:39models to build, like, robots to automate their farm, right?
0:34:42This cross-pollination and this will to create these things and learn and cross-pollinate
0:34:48across these different paths is what's really inspiring, right?
0:34:51I think earlier today I saw folks who are using AI to, like, make more robots and also,
0:34:56like, discover exoplanets, right?
0:34:58And it's that curiosity and that enablement of what AI allows you to do in lowering that
0:35:02that barrier is what allows you to be ambitious in those fields, right?
0:35:07And so when it comes to thinking about whether or not you're gonna be replaced, a lot of
0:35:10it really is the fact that you now have the skills that everyone needs in every field.
0:35:15It's not just about like writing more apps, but actually thinking about, you know, what
0:35:19were the things I didn't pursue in college to pursue software engineering?
0:35:23How can I go back to those worlds?
0:35:24And how can AI make those things easier, right?
0:35:27You can, again, develop new language and understand what makes things amazing and then go pursue
0:35:32do those things.
0:35:33And the great joy of this, and the reason why these multidisciplinary artists are so successful,
0:35:39is as they pursue new mediums, the act of being a kid again means that you have the
0:35:44naivety to ask for what might be impossible.
0:35:48The reason this cross-pollination works is if you have the judgment you've built in software
0:35:52and you pursue another field, it means that you can ask a question that people might not
0:35:56think is possible.
0:35:57And you have the courage to use AI to help you get there.
0:35:59And so there's great benefits in being able to feel new again in the field
0:36:03while having all this experience in AI as well.
0:36:07In my personal life, I've been learning the drums,
0:36:09I've been building custom drumming software,
0:36:11AI is helping me learn how to use Ableton,
0:36:13I'm connecting my drums to MIDI and learning there.
0:36:16I've been using computer use to teach me new software
0:36:19that I never really had time to learn,
0:36:21and I've even been using AI to help me design physical engineering.
0:36:25So I'm 3D printing parts, building chips, and again, these are things I was always curious
0:36:30about but never had the time.
0:36:32And again, part of having AI is not only getting my time back, everything else is accelerated.
0:36:37Right?
0:36:38And so being ambitious is about exploring those pursuits, not just spending more tokens
0:36:42like building an app.
0:36:45And so this is also what I want you guys to think about, right?
0:36:48Again, what fields did you put aside?
0:36:50How can you pursue them now that you have more time with AI?
0:36:54And so when I'm talking about, like, increasing ambition,
0:36:57I'm not just saying, like, become a biologist, right?
0:36:59What I'm really saying is that, like, part of your job now,
0:37:01to become more ambitious, is to be really inspired.
0:37:05Pursue these things. Learn that language.
0:37:07And also for that next generation, to be inspiring.
0:37:10Right? When people used to ask me how to, like,
0:37:12learn how to code, it was never a case where I said,
0:37:15you know what, like, you should go read a textbook on Python.
0:37:18It was always build something, right?
0:37:19It was always go build a weather app.
0:37:21go build a note-taking app, have a real problem and solve it.
0:37:25And now the scope of these problems are much bigger.
0:37:27And so as software developers, I think our job now
0:37:30is to not only build amazing things and share them,
0:37:32but also inspire the next generation to have more hope, too.
0:37:36Right? Because maybe software is changing,
0:37:37but really we're going to be able to contribute
0:37:39to so much more things. Right?
0:37:42And like, we see this everywhere.
0:37:43Like, for example, the Codex app and Sites,
0:37:45these are all things that were acts of love.
0:37:47Codex app, the Sites product,
0:37:49we're all built by people who really cared about these things,
0:37:51or talking to people online, and working through these things.
0:37:54Right, and now our PMs can do engineering,
0:37:56our engineering can do design,
0:37:57and this cross-pollination is inspiring others to do the same,
0:38:00and take ownership of that work.
0:38:02Right, and most of the things here that exist are here
0:38:05because people are tenacious, and are trying,
0:38:08and want things to be beautiful.
0:38:10You know, and ultimately, our mission of OpenAI is to benefit all,
0:38:14and ultimately what that means isn't just having OpenAI
0:38:17I have all these AI to control and, you know,
0:38:20use agent swarms or whatever,
0:38:21but also making sure that everyone has this technology
0:38:23so they can contribute in their own ways.
0:38:27And so that's basically it, you know?
0:38:29Think about the things that you own.
0:38:31Go out, get inspired, develop the language you need
0:38:33to describe what beautiful things look like,
0:38:35and just make sure you're actually curious enough
0:38:38and take the time to pursue these new interests.
0:38:41And then that's how you become more ambitious.
0:38:43So I'm really excited to see what you guys are building
0:38:45and come by to the booth to chat.
0:38:48Thank you.
