Event Replay: Codex is for Everyone: Why Codex Matters Beyond Code

[00:00:00] Chris Nicholson: Good afternoon. Welcome to OpenAI Forum. My name is Chris Nicholson. I'm with the Global Affairs Team, and I'm glad to be here with all of you. So the Forum, as some of you know, is a place where we talk with experts about how AI is being used in the world. Today's conversation is about Codex and why it matters beyond software engineering. More and more, people are using Codex to help with knowledge work and personal work. They're using it to reduce friction, perform tedious tasks, or just to understand a problem, organize information, and plan a document that they can share. So that makes Codex helpful for researchers, educators, operators in industry, small business owners, leaders, knowledge workers of all kinds. And today, I'm joined by Thibault Sottiaux. Thank you for coming.

[00:01:02] Thibault Sottiaux: Of course.

[00:01:03] Chris Nicholson: Thibault is the head of Codex at OpenAI and he's going to help us understand where Codex came from, what the team is learning from people using this tool, especially non-technical users, and how folks can begin to use it outside of coding. So, Thibault, great to have you here.

[00:01:18] Thibault Sottiaux: Great to be here.

[00:01:19] Chris Nicholson: Yeah. Well, let's get started. To ground this, Codex was created as a tool for developers. Maybe you could tell us when did Codex first launch, and what problems were you and the team trying to solve at the beginning?

[00:01:31] Thibault Sottiaux: Yeah, so the journey started a long time ago at OpenAI. We've always been fascinated with helping, just figuring out how we can create very useful models that would then help us also accelerate development. One of the grand challenges for a while has always been coding. Being able to code at the level of a prolific software engineer was a grand challenge, like roughly two years ago when we started on this journey. The first public version of Codex was something that we now refer to as Codex Web. It was this idea that we would have an entity running in the cloud, accessible over a web interface. You would put your task in there, and it would go and look at the repository, like a code repository that you have. It would figure out what the right set of changes were in order to fulfill the task, and then it would open a pull request on GitHub. So, it was like fully isolated, fully packaged. You would just save your intent, and then you would receive a code change at the end. That was roughly a year ago.

[00:02:39] Thibault Sottiaux: What we found was it was a little bit too high friction. It was too hard to set up, and then also, a lot of developers just have a beautiful setup on their own computer. You needed to go and reproduce all of that on our own cloud environment setup. So, the friction was too high, and the models were also not at the level where they would get it right all the time. It was hard to iterate with the model, and we shifted our approach after that.

[00:03:09] Chris Nicholson: Yeah, great. So, you were building a tool for yourselves, and it sounds like there was a pivot to where you realized that it should work on everybody's machine locally.

[00:03:17] Thibault Sottiaux: That's right. We realized that doing this on our own cloud machines and asking everyone to configure it there was too difficult because the models were not yet at the level where it was reliable for long-term horizon tasks. It was just not working.

[00:03:31] Chris Nicholson: When did you first see people using it for tasks outside of coding?

[00:03:36] Thibault Sottiaux: That's been more of a recent shift in the last six months, I would say, ever since we released GPT-5. We've had that step change in generality and reliability for general purpose tasks, I would say around 5.2 as well, where it became more reliable for long horizon tasks. One thing to realize as well is that even for software engineers and technical people, the majority of the day is not actually coding. The majority of the day is looking at tickets, prioritizing, having discussions around how we should actually solve problems and deciding on the architecture, solving bug reports, investigating bug reports, and figuring out whether it's actually a user problem, where the problem actually is, dealing with outages, being on call. In general, software engineers spend maybe 20-30% of their time actually coding. What we saw is a lot of the people who had adopted Codex were already using it for their day-to-day work. It was obvious that we were holding a technology that was much more powerful, and that we wanted to make useful at scale for the entirety of society.

[00:04:58] Chris Nicholson: So it sounds like, to be a software engineer, you have to do a lot of things that aren't coding, and people were dogfooding. They were using their own tool to do these non-coding tasks.

[00:05:08] Thibault Sottiaux: In order to be useful at coding, we were thinking, okay, you need access to more context. Maybe if you have access to all of the information that we have in Notion, information that you have in documents, then actually we saw that the agent would become more useful at solving the task. And then we were trying to just make that more and more reliable up until the point where now the majority of tasks that are being performed in Codex are actually non-coding tasks.

[00:05:40] Chris Nicholson: That's really interesting. So first of all, Codex was searching code, and then you realized it's also useful to search a bunch of documents and come back with information. And obviously, that's something that all of us doing knowledge work have to do.

[00:05:54] Thibault Sottiaux: That's right.

[00:05:56] Chris Nicholson: So maybe, do you remember the first moment when you were like, oh, this is actually going to be for everybody? When the ambition changed.

[00:06:06] Thibault Sottiaux: Yes, we were in the middle of a launch for Codex and Alexander Embiricos, who is the lead product manager on Codex, was using Codex to keep track of the state of all of the changes that we were trying to land in the run-up to the launch. I had never seen someone be as productive as Alexander at that moment. It seemed like he had many little Codex agents doing his work on his behalf, chasing people up and then updating a document with the latest state of things, like, oh, you know, we actually still need to publish this specific feature. We have this user feedback, we have that. It was sort of coalescing all of this information from all the user feedback that we have, the information from the developers, and keeping a plan neat and up to date while Alex was having a discussion with me. I felt like, wow, you know, we're really changing everything here, not just software engineering.

[00:07:06] Chris Nicholson: So in the world before Codex, Alexander would be either himself digging through Slack channels or documents or GitHub PRs and just like everybody, spending so much time with coordination, and then you saw he had somehow delegated that really time-consuming work to these tools, and they're doing it while he's in a meeting with you.

[00:07:24] Thibault Sottiaux: That's right. Our models are very good at gathering the right context and summarizing as well. And so that's a strong use case that we see, and he was doing that and then also using it to chase up information. So we have it connected to Slack; it's able to send messages, the Codex agent can send messages to people to also ask the Codex status on this thing. So doing all the chasing on behalf of Alexander.

[00:07:52] Chris Nicholson: Chasing takes so much time. So there's Alexander, he's one of the most productive people you've ever seen. He's using Codex in this way. What's the ripple effect? In concentric circles around the people around him, the team, what gets shipped?

[00:08:03] Thibault Sottiaux: The ripple effects that we've seen as engineers have been greatly accelerated, and we are now building faster than ever. Adjacent jobs are changing, like the role of the designer is changing, the role of the product manager is changing. And then, you know, we looked at how can we accelerate and give them also the means to be more productive. Then we saw bottlenecks shift to communications and marketing, where we were suddenly producing so much work. It was a challenge to sort of tell the world and keep the story coherent. So the bottleneck was shifting to other departments at the company.

[00:08:50] Chris Nicholson: Yeah. So OpenAI is shipping really fast. Would it be possible without Codex?

[00:08:57] Thibault Sottiaux: No, I don't think so. At this point, Codex is critical for us.

[00:09:00] Chris Nicholson: And do you think other companies, maybe with ten times more engineers, do you think other companies are similar enough to OpenAI and like the problems they're trying to solve for the world that they can follow a similar playbook?

[00:09:08] Thibault Sottiaux: Yes, I think now the state of the technology has reached a point where these agents are capable of very general work. There aren’t really many things that you can do behind your computer that the agent isn't able to also help with. A lot of the work that needs to go into preparing a presentation to align stakeholders, for example, gathering context on public perception, doing some marketing research, organizing. It's used a lot in our finance department. Sarah Friar talks about this a lot, but she organized the latest fundraise with Codex, with the help of Codex. It was a great accelerant there as well. And so there's this beautiful generality to it, and it's evolved much more to, like, it's not for generating code, it's just to perform general-purpose tasks really.

[00:09:56] Chris Nicholson: So you mentioned that roles are changing; what kind of changes do you see?

[00:10:01] Thibault Sottiaux: The change that I see is that everyone has to deal with things moving much more quickly and adapt much more quickly as well. Hard problems, things that used to take days now maybe take like a couple of hours across the field. We're seeing this in science, we're seeing this in engineering, but we're also seeing this of like if, you know, I would have expected like doing an in-depth marketing research or analyzing the public sentiment around a new feature, you know, that would have taken a lot of time to go and find all the sources and summarize and then condense it in a way where different people could consume that. Now that's really being compressed to like a couple of hours of work and being completely automated.

[00:10:48] Thibault Sottiaux: Everything is suddenly moving much faster as well, and you need to adapt to things, you know, just the pace having picked up. People are able to do much more by themselves as well, and there's something delightful about that too; it's very empowering. So maybe before, you're like, "Oh, I need to talk to someone about this, I don't know," you can just do it. We're seeing this a lot on data questions within companies, where it's like you have a question about, "Oh, you know, how successful are we in this market? Are we growing in India? What's happening in Korea?" And then now everyone is just empowered to ask the question, like, "Hey, you pull up the dashboard, I don’t really know where it is, but I'll go figure it out," and then you can go and dig into all the details of the business, and you don't need to go and talk to data analysts. You can leave them alone; they can do more interesting stuff than answering your little questions.

[00:11:42] Chris Nicholson: Yeah, I remember the days when data analyst teams had long queues because they're the shared resource of things they owe to all the other teams. And as long as other teams have access to the underlying data, very often they can extract insights by themselves. What I feel like we're in is this historical moment where it used to be that the people who had problems and the people who built solutions were two separate groups of people, and the conversations they had about the product were really long. You eventually got to something that was kind of okay, but you couldn't push it further because nobody had time. And now, the person who has the problem can build the solution very quickly and iterate through the changes they need.

[00:12:17] Thibault Sottiaux: Yeah, that's what I see with designers as well on the team. They're pushing code, they're making changes, they're in there with the developers shaping the product as they want to, to kind of bring the best out of it and create an amazing experience. They don't have to go and convince the engineering team to prioritize what the engineering team probably feels like are fairly minor changes. But for our amazing designers, they're like, "Oh no, this is just really elevating the craft and the experience for millions of users out there," and so it's very empowering to them too.

[00:12:57] Chris Nicholson: It feels like we're kind of in this era of home-cooked personalized software on so many levels.

[00:13:08] Thibault Sottiaux: Yes, it feels that this is the wave that's coming where everyone will be able to have their own personal software and maintain it. And it just does exactly what you need it to do.

[00:13:17] Chris Nicholson: Yeah, you were telling me before we started about something you built for yourself that visualizes some data. Do you want to share that with us?

[00:13:26] Thibault Sottiaux: Yeah, sure! This is just really a fun exercise. I have it on my screen here and I live in San Francisco, so I think the price of bread is outrageous in San Francisco; I moved from Europe. And so I was like, "Hey Codex, can you find the best loaves in the city and create me a spreadsheet so that I can understand where are those loaves, where should I buy them, and what's the associated price?" And so it went, and it worked for five minutes, and it created this spreadsheet which you can see here.

[00:14:02] Thibault Sottiaux: Can we see that? Yeah, everybody see that? Great! And so you can see you have Jane the Bakery, you know, Arsicault, it has found the different loaves, a little description, and the price. So it seems like if I want to go and find a good loaf of bread, the sourdough loaf, I should go to Neighbor Bakehouse, which is in Dogpatch. And then I was like, "Okay, it's great, I have this spreadsheet." I could have iterated on the spreadsheet and asked for more information, but instead, I wanted to consume it in a bit of a more visual way. So I said, "Hey, how about you create me a web page?" And so it just put all of these on a map here. I can just, and this is just really personal software that was created for me, and this took four minutes for Codex to just generate. I can go and click around, and I could probably ask it to show me a picture of these loaves of bread as well.

[00:14:54] Thibault Sottiaux: And this was just done in ten minutes. I didn't even look at it, it was just a single prompt for me. And then the way that I like to engage with Codex, I would like to show you, I do everything over voice. So, hey, I'm in San Francisco, I'm really into bread, I want a map of all the loaves that I can buy, and just restructure it with a little explanation of where I can buy it, and the price of the loaf. And then I don't even need to type anymore, I'm just going to go and do that. And then I can go and just do other stuff on my computer. So, this is not a thing where I'm even actively spending attention on it.

[00:15:40] Chris Nicholson: So, you're saying anybody can do this, if they care about some data, and they have access to it, they can basically make a website, analyze it, visualize it, share with their friends, family, co-workers.

[00:15:53] Thibault Sottiaux: And we used to talk about writing code, it seems like such an anachronism, now you're kind of speaking code, right? It's using code under the hood, but it's more of a tool for the agent. You don't even need to know about it. It's editing and creating spreadsheets and documents and slide decks for you. It's creating websites for you. You don't need to understand how it actually performs its work under the hood. It's just using code as a tool, and this is why there's this beautiful generality to it.

[00:16:25] Chris Nicholson: Yeah. So, in the old days, if you had the skills, this might have taken you a weekend to create something as polished as that.

[00:16:31] Thibault Sottiaux: Oh, it would have taken me quite a bit of time to go and find all that information.

[00:16:34] Chris Nicholson: Yeah, right. And that was just an instant. And then if you don't quite like it, you can just tell it to change it, right?

[00:16:36] Thibault Sottiaux: That's right. If I'm like, oh, actually, I wanted to do the same, but for coffee. Hey, do the same analysis for coffee, not just bread. That's it. That's all I need to do. It will go and work for eight minutes and likely it will end up with a website just about coffee in San Francisco.

[00:16:58] Chris Nicholson: And you're looking for, you're optimizing for price and quality, right? It's actually more complex than just give me cheap bread.

[00:17:04] Thibault Sottiaux: Yes. So you can express your own preferences. Codex will just go and take that into account and just really work to satisfy your prompt.

[00:17:14] Chris Nicholson: Right. So it gets the data, it visualizes the data, it gets you to an insight about the world. Then you can make a decision to suit your needs. Which is kind of like the universal loop we're all in in our lives. How do I actually decide, make a better decision to reach my goals?

[00:17:31] Thibault Sottiaux: Yes. And this can be very simple. Can be as complicated as the same process was used for the latest fundraiser opening. So Codex is capable of the smallest things and then the most complicated things as well.

[00:17:45] Chris Nicholson: Yeah. Amazing. So are you now using it like Alexander? Or are you kind of coordinating?

[00:17:52] Thibault Sottiaux: You can see here in my sidebar, this is before we started to talk. I fire off a little task all the time, maybe over a hundred different tasks per day that I just hand off to Codex and it just goes and does it for me. This goes from organizing my desktop files, managing the compute fleet, helping me understand how the on-call rotation is doing and how the engineers are doing, helping me understand the launch schedule that's coming up, and flagging anything that's at risk that maybe I need to look into.

[00:18:29] Thibault Sottiaux: I use it as a little chief of staff and I run this every day. For example, you can do a little automation and go through Gmail, Notion, Calendar, and flag, and give me a summary of my day and flag anything that is at risk. And then I can say, okay, do this daily at 9 a.m. And then I can just go and schedule that and then it will run it and then I will find it in my inbox at 9 a.m. every day.

[00:19:07] Chris Nicholson: So it's like you've got your own chief of staff. Hey, don't forget to do these things. It's even helping you prioritize.

[00:19:15] Thibault Sottiaux: And that's right. It's helping me prioritize. It's helping me spend my attention where it matters the most. And it's helping me also by I don't have to go and do the small things that would otherwise just take me time and maybe I would never get to.

[00:19:30] Chris Nicholson: What were the things that annoyed you? The most tedious things in the pre-Codex time that you spent hours every week on?

[00:19:40] Thibault Sottiaux: The thing that annoyed me most was I would not do certain things because I felt like, oh, I don't have the time to do it myself, but I need to go and annoy someone to ask for this information. But I felt like it was maybe not important enough to go and put this on someone's desk and be like, hey, can you help me with this question? So now I feel like I can just get the information that I want, have the little reports that I want. I can build all sorts of personal software that I didn't have the time to do. So it's just like a really, it gives me joy because it is doing all the little tedious things that were very manual on my computer before. And now I'm like, OK, I'm just in the Codex app and doing all these things for me. I can just focus on the things that I feel I actually want to think about.

[00:20:30] Chris Nicholson: Yeah. So I know people often talk about it as like this would have taken weeks. Now it takes seconds, but it actually feels like this never would have happened. It would have taken infinity time because we never would have gotten to it. And now it's, there's a lot of that.

[00:20:46] Chris Nicholson: Are you enjoying your job more because these things are getting done that used to annoy you?

[00:20:50] Thibault Sottiaux: Yes. Yes, I am. There's so much that I don't, you know, like for example, every day I'm like, oh, what's the latest news? And it has very personal instructions for the kind of news that I care about. And so that is an example, like I would have never received such a report about the news. And it gives me great joy to just read this in the morning and feel a little bit more informed. A lot of things on when it comes to investigating, for example, we have a bug report from a user. I see it on Twitter. And I was like, oh, you know, that is a bummer because I would have usually put this as like this is a low priority thing. This is probably affecting like three users on the planet. This doesn't feel like a thing. Now I'm like, oh, let's just put this in Codex and it will get it done. And I can feel much better because I don't feel like things are falling through the cracks.

[00:21:54] Chris Nicholson: It's reducing the cognitive load for me. It makes me think that this is actually a tool to address burnout and information overwhelm. Like people, we're all kind of surrounded by tools that are supposed to help us. But again, we're kind of trapped within our tools. And this is a thing that's kind of liberating us as a tool and maybe like so many people, teachers, doctors, so many people feel burnout and overwhelmed because they're trapped in those tools. Manual data entry, that sort of thing. Do you feel like it almost feels like our relationship to tools is changing as a fundamental shift in how we relate to tools?

[00:22:30] Thibault Sottiaux: Yeah. To me, that's the promise. The promise is you have a trustworthy partner, almost, that you can go through. And, you know, we'll do a lot of work on your behalf to a level of trustworthiness where you're like, hey, it's like if I delegate something to you, you'll get it done. If you don't get it done to the level of satisfaction, you'll flag it to me. But I know that it will be done well. I know that it will be done maybe to a level of quality that's superior to doing it myself. I can also trust this partner to shield me from a lot of the noise and to flag important things in a timely fashion. So I don't have to go and consume information in like seven different apps and be bombarded by things that I don't actually care about.

[00:23:25] Thibault Sottiaux: What I imagine the future to be like is I don't even have to read my email. Maybe I just have my little personal agent that is just reading my inbox for me, flagging me whenever there's something really important, asking me for input, and then just doing the work.

[00:23:45] Chris Nicholson: So you don't have to search for needles in a dozen different haystacks.

[00:23:49] Thibault Sottiaux: Yes. The needles are kind of arranged in a newsletter.

[00:23:55] Chris Nicholson: I can say like here are my goals for today, help me out, manage everything else, and trust that that will happen. That's amazing. You must have seen over the past few months what you can trust this tool with is changing. You mentioned time horizon, the time it takes to do a complex task. That's changing.

[00:24:11] Thibault Sottiaux: Yes. So we actually launched a new thing there. There's more of an advanced feature, which is slash goal. So in Codex you have slash commands, which allow you to, for example, enter different modes. Slash goal allows you to enter a mode where you give a long-term goal to Codex, and it goes and is really relentless at it. And so, for example, you can give it the goal of solving a very hard mathematical problem, and it will go at it for hours, maybe days, maybe weeks, until it has decided that it has satisfied the goal.

[00:24:50] Thibault Sottiaux: We're seeing it being used to do amazing things on improving the performance of programs, rewriting entire programs from one language to another. It's used in science problems as well. I've seen really cool breakthroughs on mathematics and physics breakthroughs with it. And this is fascinating to me because maybe a couple of months ago, we were, oh, you know, this is exciting. It's working for 10 minutes. Now we're talking about agents working for weeks on the hardest tasks.

[00:25:20] Chris Nicholson: Sometimes I think genius is just the ability to think about the same thing longer.

[00:25:25] Thibault Sottiaux: Yes. Maybe if you have infinite time, probably a sufficiently smart human would be able to come up with everything that we have done in humanity so far, so that feels very interesting.

[00:25:39] Chris Nicholson: So goal, you can use that in the CLI. Can you use that in this Codex app that you're showing us right now?

[00:25:45] Thibault Sottiaux: Yeah, it's coming. It's not yet launched, but it will be available.

[00:25:50] Chris Nicholson: OK, we'll let the community know when goal is available on all surfaces. OK, awesome. So it's running for days and it's just going, going, going in the background, and it comes back and reports to you, hey, I'm done or I'm stuck on something?

[00:26:04] Thibault Sottiaux: Yeah, that's right. And where the future is headed, I think, is where it's not going to stop, where you're just going to have an agent that runs like 24/7 and does useful things for you and gets steered along the way.

[00:26:19] Thibault Sottiaux: Right now it's very turn based where you go and you have a specific task and you're like, oh, I want you to go and solve that task. Go find all the loaves in San Francisco and tell me the price. That's great. Solving tasks is super important. Goal oriented task solving is also a big unlock in terms of the value that you can create for you. But the next step, I think, is it just runs continuously and it does useful things, whether you instructed or not. And then maybe at some point it's like I've done all the things that I think are useful. I'm just going to sleep for a little bit until I talk to you. But it's not like it's just working when you have a very clear instruction.

[00:27:02] Chris Nicholson: Is there any trick to specifying the goal you want to? Like, how do you make goals succeed?

[00:27:09] Thibault Sottiaux: Yeah, so one of the shifts with agents compared to, you know, maybe if you haven't experienced, if you haven't installed the Codex app and it's the first time you're using it, just explore a little bit. You can engage in a very chatty way about what it's capable of. It understands its own capabilities. So you can just ask it what it can do. But a good trick is to be precise on helping it evaluate its own success. And so if you're able to describe what good looks like and what's solved looks like and what you want to see at the end of the completion of the task, then Codex will be able to understand whether it's doing well and whether it has completed the task.

[00:27:59] Chris Nicholson: So can you put a number on it? Can you be precise?

[00:28:03] Thibault Sottiaux: Yeah. So, for example, you can say, OK, I'm presenting my work. I would like a slide deck. I would like it to have 10 slides. The first two slides, I want it to contain this kind of information. The next six slides, I want it to just really go into the meat of the problem and do a technical breakdown. And then I want two slides at the end with open questions and Q&A. And if you are clear about it and you're specific about what you want as the output, then it's much more likely to succeed.

[00:28:39] Chris Nicholson: So very similar to what you might do with an assistant or a colleague.

[00:28:39] Thibault Sottiaux: That's right. Yeah. That's right.

[00:28:40] Chris Nicholson: OK, so we take questions from the community. Let's do it. And I'm going to read some of those. So we've got one here from, and excuse me if I mispronounce this, Gauging Karya, principal engineer in financial services, is asking what would make non-coders move away from ChatGPT?

[00:28:55] Thibault Sottiaux: I think it's going to be an evolution and it's really, I don't expect that everyone will want to move to Codex, but it is a nice complement to chat. What I recommend to use it for is to do things for you. So anything that you know where it's a manipulation of files and your computer, running things on automation and doing things in the background every couple of hours, you can do that with Codex. And for me, ChatGPT is still my go-to for quick answers.

[00:29:47] Chris Nicholson: Yeah, I remember the old days for me. I would be copying and pasting code from ChatGPT into a file or a terminal, but you don't have to do that anymore. The copy and paste era is over. Codex can just work on the code and the data.

[00:29:58] Thibault Sottiaux: Yeah, and if you have a file on your computer, you have pictures on your computer, just tell Codex, like, hey, use this file, read it, and it can just do it. You don't need to go and click on it manually.

[00:30:14] Chris Nicholson: Okay, we've got one from Anastasia Chweva. What do you think becomes the biggest bottleneck in enterprise adoption, and then the choices are model capability, human trust, or organizational process design?

[00:30:32] Thibault Sottiaux: I don't think it's a capability thing. The capability is there. That's not what's holding back adoption in enterprise. I think it's primarily trust, and for trust, it's about is it safe and is it secure. If you have an agent just run around in your company and maybe delete sensitive files, or upload information and send an email that contains and exfiltrates information that it shouldn't. People aren't gonna use that. That's terrible. So, we've been thinking a lot about that. By default, agents run inside of sandbox with tight controls. You know exactly that they are able to only access a specific portion of the file system. And so, you can constrain it to a folder. You can disable network access. We're giving a lot of enterprise controls. I think this is very key to continue to expand from there. There's also investments that we've done. It's on our alignment blog post. We call this feature auto-review, where we have created an agent that is capable of reviewing the actions of the primary agent. And the way it works is that the primary agent, Codex, is incentivized to do the work for you. At times, it might take an action that is maybe a little bit risky. And so you have another agent that reviews every action that it's taking and flags and stops it if it's a high-risk thing. And I expect more innovations like this will be necessary.

[00:31:58] Chris Nicholson: Like a referee or an umpire, that's against the rules.

[00:32:03] Thibault Sottiaux: Yes, yes, and be like, hey stop right there. This is too risky of an action. I would like you to take a less risky action.

[00:32:09] Chris Nicholson: So just to help the non-engineers in the audience, a sandbox, is that just like a folder in a file system? Like do you say don't go outside of this folder, just do things within this?

[00:32:19] Thibault Sottiaux: Yeah, so that's right. It's similar to how you can think about if you're in a company like, maybe you only get access to information for one part of the organization that you're in, and people have different silos. This is like the equivalent of a sandbox for an agent. You can restrict what it can have access to and the set of actions that it can take.

[00:32:41] Chris Nicholson: Yeah, and so a rule might be, you can read this data that IT gave you access to, but you can't write back to it, you can't delete it, you can't change it, the data is preserved, but you can kind of extract and analyze it.

[00:32:57] Thibault Sottiaux: Yeah, for example, you could say, oh, you can only have read access.

[00:33:00] Chris Nicholson: Cool, okay, we've got one, Jason Deluca, owner and principal consultant, Crossing Point IT Solutions. What are non-developers doing that makes Codex work well for them and what habits separate the people getting results from ones who feel stuck?

[00:33:15] Thibault Sottiaux: I think what I've seen is engaging with creativity and talking to others who have seen success. I am surprised myself at some of the use cases. So being part of a community is great. The second thing is trying to engage with it with precise instructions instead of being quite vague about what you want. And so putting some effort into just really being like, hey, this is precisely the outcome that I would like.

[00:34:01] Thibault Sottiaux: We talked about this a little bit, but I would just really consider agents to be someone new who just joined a company, has no context about you, doesn't know what you like, doesn't know about your preferences. How would you engage with them? You would probably give them quite a bit of information about where to find things, how to think about relevant documents to read. And so often people forget about that. And then the third thing is connect a lot of different sources. So we have plugins for almost everything now. We have more than 100 different plugins. You can connect your calendar, docs, Notion. You can connect your favorite tools. The more you give it access to your tools and information from your world, the more useful it will be.

[00:34:46] Chris Nicholson: So the more and better context gives you more and better results. And some of that context is in your head. And so it's your goals, it's things you've experienced that are not in the docs, and you gotta be a good boss and actually share some of these things.

[00:35:02] Thibault Sottiaux: That's actually interesting. So I have taken the habit of writing everything down. So a lot of my own thoughts, my goals. It's actually in files on my computer. And Codex has access to it as well, so that it can align itself a little bit more. So, obviously, yes, if it doesn't have access to your brain, it can't read your mind. So you have to verbalize that sometimes.

[00:35:27] Chris Nicholson: Yeah, neat. Okay, we've got one more from the community. Daniel Green, Kansas City AI Collective. What are your favorite real-world uses of computer use so far, especially outside traditional coding?

[00:35:41] Thibault Sottiaux: I use it for shopping. Where it will go and order stuff for me. What I do is, on my personal computer, I have Codex do meal planning. And then it will go and actually order the ingredients. So that's the thing that I use it for.

[00:36:03] Thibault Sottiaux: Another thing that I've seen people use it for is that it's almost impossible to find which settings popup to open up, be it on Windows or MacOS, to do something on your computer. And so just being like, "Hey, Codex, I'm trying to change this on my computer, can you show me how to get there?" And it will go and take you around and click on the right things. And then, you know, you learn something about your computer. Sometimes, as well, there's something that you just kinda want to adjust. Some slides or integrate an image. And then computer use is great for that, as well. For technical people, it is very, very useful for QA, where Codex can then open the app and click around and test that it actually functions.

[00:36:58] Chris Nicholson: Yeah, amazing. Okay, let's see, what do we got? Sai Srihamsini N, AI/ML Engineer at AdaptiveView, what are the biggest mistakes people make when prompting Codex? I want to know the answers to that one.

[00:37:10] Thibault Sottiaux: Oh, um... As you progress in delegating to Codex, it can be, at some point I see it becoming very tempting to just delegate everything, including your own understanding. And not using Codex enough to elevate your own understanding of the problem. If you're just purely like, "Oh, I'm gonna delegate everything," at some point you might realize you don't actually understand what's happening anymore. And then you're not as grounded and you're going to lose a little bit of your productivity and that's not good.

[00:38:01] Thibault Sottiaux: So spending more time to also gather information and have it explain things to you. It can draw diagrams, like Images v2 is amazing at this because it's very good at rendering text too. And so I see people use it all the time to go and read our plans for launches or things for marketing or parts of the code base and then create images to just explain a concept to you if you're trying to learn something. So I would say the mistake I'm seeing is just too much delegation and not enough using it to just help you understand stuff.

[00:38:37] Chris Nicholson: I totally agree, whoever does the work does the learning, basically. So I built myself a little skill, a tutor skill, and it flips the tables and it asks me to give it answers about what it's tried to teach me already, so I can force myself to do that learning of active recall.

[00:38:54] Chris Nicholson: Yeah, super cool. Let's see, what else do we have? Why does Codex matter beyond software and where do you see these use cases going next? So we're really talking about the future. We've talked a lot about how things are changing up until now. I think we're the future for a lot of people but there's a future for us too and for everybody together.

[00:39:10] Thibault Sottiaux: Yes, so fundamentally what we're building is a very general and very powerful agent that if connected to the right sources of information and with increasing access to all the information in the world and given the ability to take actions upon the world, we'll be able to do pretty much anything that you allow it to do. And so the promise there is just incredible value creation. A lot of things that would have taken prohibitive amounts of time you would never have gotten to or just too hard to accomplish for humans will become possible.

[00:39:52] Thibault Sottiaux: Everyone, we will distribute this. The intention is to distribute this as broadly as possible and to give the entire world access to these capabilities and these agents. And so it's just really about lifting the amount of things that people can even dream about accomplishing.

[00:40:12] Chris Nicholson: Yeah, I've spoken to everybody from prize-winning physicists to industry project managers. They've come to me and said, I had a stack of good ideas and I've been able to start work on them, and they never would have been realized in the world before. It feels like we're entering a golden era.

[00:40:29] Thibault Sottiaux: It feels like it.

[00:40:30] Chris Nicholson: Yes.

[00:40:31] Thibault Sottiaux: Yeah. Amazing.

[00:40:32] Chris Nicholson: Okay, well, this was a great session. Thank you.

[00:40:35] Thibault Sottiaux: Thank you.

[00:40:37] Chris Nicholson: And I just want to follow up. Tibo mentioned that having a community will help you learn how to use Codex better. I just want to make the point that OpenAI Forum is a community where people are using Codex and we can teach each other and support each other in learning that better. I hope everybody in the session today will continue to engage because it is a place where skills are modeled and transferred so that you can bring this tool into your life in new ways.

[00:41:00] Chris Nicholson: The question we tried to answer today is why, why should people who don't care about code care about Codex? I hope we answered that. I also like to make the point that although the word code is in Codex, codex actually means book, which is something we all are used to. Right? So, Codex is a more general word than just code. But I hope this conversation made that answer concrete.

[00:41:24] Chris Nicholson: Codex is useful for developers, also useful for people, anybody, knowledge workers, people working with information, you know, searching for the information they need, the needles in the haystacks, struggling to make sense of it. Data analysis, visualize it with kind of instant visualization, like Thibault showed us, and prioritization and performing complex tasks in the background of their lives.

[00:41:50] Chris Nicholson: So we think that Codex is bringing those powers, those superpowers to lots more people. We're really excited about what that's going to mean for our economy and society and their businesses. As we close, I'd like to ask everybody listening to think of one workflow in their life or in their organization where Codex might help.

[00:42:08] Chris Nicholson: And that could be a research brief or a plan or an onboarding process or reporting or a decision memo. Just give it a try, because you've heard a lot from us today, but the clearest message is just going to be your experience with the tool. You don't have to trust anybody else about it. You can just experience that tool. It's there, available to you.

[00:42:30] Chris Nicholson: So thank you again, everyone, for being part of OpenAI Forum. We appreciate your time and your curiosity and the thought everybody brought here and especially the folks who got their questions in. Those were great questions. So thank you.

[00:42:45] Chris Nicholson: We have an upcoming event. It's Tuesday, May 26th, How AI Could Detect and Prevent the Next Pandemic with the Gates Foundation. Super excited about that. So thank you, everyone. And thank you, Thibault.