Fidji Simo on GPT-5.2 and OpenAI's plan to become everyone's personal super-assistant

seconds. Our next guest is Fiji Simo from OpenAI. She is the CEO of applications. [music] We are very honored to be joined by Fiji. Welcome to the show. How are you doing? Great to have you on the show.

Great to be here. Thank you so much.

Thanks so much for joining. Um, I would love to kick it off with a little bit of uh a little bit of backstory on just what was the conversation like uh to recruit you to have you join OpenAI? It's a big move. It it rocked the timeline. Um, and I'd love to know what was the pitch. What was the opportunity? How did you wind up at OpenAI? Well, I had the extreme privilege of being on the board of OpenAI for many months, like I think about a year before uh Sam approached me and uh and therefore we were already working really closely together. And then at some point uh he told me, "Hey, it looks like you know there's a lot to do and it looks like you have ideas on how to do it. So is there any chance that we could make this happen?" And I had, you know, a big job running Instacart at the time. So, it took some time to transition, but I was incredibly inspired by the mission of OpenAI already, and it felt like a no-brainer.

And uh can you take us through the the the GPT 5.2 announcement? I I mean, one of the first documents that I think uh everyone knows you for in the OpenAI context in the in the context of your new role is the the different areas where you see GPT 5 and and OpenAI having impact, Chat GPT having impact in uh folks life. Um, can you explain to me how you think those those different sections are going? Um, what excites you about the news today and kind of contextualize how do you think about the map of what uh CHACHT as an application is touching?

Yeah, that's a big question. So,

sorry, we can break it down one at a time, but

no, I I've got it. Um, you know, when I when I look at the opportunity ahead of us, I think AI is the greatest source of empowerment for people and I want CHBT to be the personal super assistant that allows you to advance every part of your life. So the manifesto that you are referring to to me was like putting on paper this notion that if you look at how the world works right now, wealthy people have a lot of support staff. They have a travel agent. They have a personal shopper. They have a financial advisor. Imagine if we could give that team of support to everyone on earth. And fundamentally, that's what I want to build. And I think with CH GPT, we have the opportunity to graduate from what is still fundamentally a chatbot that you ask questions to to a personal super assistant that gets everything done in your life. Um, when you put that in the context of uh the announcement today, we're super excited to have launched 5.2. It's the leading model on on uh kind of [laughter] I was really hoping for that. It's a leading model in terms of everyday professional work. It it's a big step change in intelligence. you see that on the benchmarks obviously but that gives me a lot of hope that we can be closer to achieving a lot of leaps on on these use cases. So for example if you take something like travel you have to like you know be good at long horizon task you have to understand the goal of the users very well and so with GPT 5.2 um and like the new capabilities we're getting one step closer to making all of that possible.

Yeah. So, uh, I feel like some people have a little bit of benchmark fatigue. They, you know, everyone has their own prompts and they feel like stuff's getting better, but it's hard to like, like everyone wants the original chat moment again. So, what have you done for me lately syndrome? Um, but I'm interested putting all of that aside, I want to know how are you measuring success internally around the launch of a new model? like what does the what does the the the rigor around developing KPIs obviously uh the company is in the interest of driving subscriptions and revenue but there must be interim steps and AB tests and I mean this is your lineage this is what this is what makes you who you are um how do you think about uh assessing um the impact of new models when they go out aside like like downstream of the benchmarks in the actual application

yeah so that's like my entire job is turning the research breakthroughs into products that people use. And so for example on 5.2 we just talked about the consumer side. I can go back to that but on enterprise for example it's really important to us. We serve a million businesses uh across the world. Uh what we look at is like can companies start to do things that they weren't able to do before. We just released earlier this week an uh AI report for enterprise and it said that 75% of people were saying of of workers were saying that with AI they can accomplish tasks that they couldn't do before and then we see that even on the coding stat on coding if you look coding uh quer messages from non engineers has grown 36%. And so I look at all of these things as like these are all like the research team gives me these like magical superpowers. My job is to put the superpowers in the hands of people and and at at home and at work and the thing we measure is can they do more and are they actually like understanding the use case and doing more and there's a lot you know there is a big narrative around like is AI adoption really happening is it going fast enough but what I am seeing is that there is a lot of pent-up demand as long as you cross those capabilities so I'll give you an example all companies that I meet with tell me that if there was a way to generate spreadsheets much like faster with AI, generate slides much faster with AI, they would use it for a lot more knowledge work because you know all all knowledge workers generate spreadsheets and slides. With 5.2 like that's a big step up in the ability to do that. Like I can tell you with 5.1 I was trying to do it and the spreadsheet was kind of janky. Things were like not not really in the right place. With 52 it's like great formatting. it totally understand what like you know the numbers are about it structure them the best way and these are the things that like help unlock completely new use cases and ultimately growth for the customers that we work with. Do you do you feel like there's a there's some sort of fundamental cultural difference between um you know your previous roles where there hasn't been so much of this divide between the consumer side of the business and the enterprise side of the business on day one. I feel like chat GPT is unique in that it's used by kids and college students and adults and people use it just they bring it to work with them alongside and then enterprises and the government's using it. Uh whereas for something like a social media application, it didn't really have that same oh we need the in we need this product to work for us in an enterprise context necessarily. Um how how has that how has that shift been designing an application that can be used? It's sort of the same underlying model, but it's used in such different ways.

Yeah, it's a great question and that's that's the thing that we spend a lot of time working on. This thing that's really interesting is that part of the reason why we're winning in enterprise is because we're winning in consumer. When you talk to CEOs, they tell you, well, if my workers are already completely familiar with the technology because they use it for their personal use, that's much easier to deploy that in enterprise and have that be adopted. So that familiarity is actually really helping us. And so the thing we're trying to do is really keep that familiarity but upsell the tools that are really specific to enterprise. So I'll give you an example. Uh connectors is a is a great example. We do have connectors on the consumer side. You can connect your chat GPD to your Gmail and and things like that. But on the enterprise side, it's even more critical. You have to be connected to like Slack to like all of your enterprise knowledge. And so these are cases where we push harder on the enterprise side to make sure that companies are doing what they need to unlock the value.

Yeah. Um I'd love to hear your take or you how how you're thinking about the partnership with Disney. Uh Jordy and I were just going back and forth on it. Landed in an extremely excited place. Seems like a fantastic partnership. I think that uh I think that in general uh looking at the timeline this morning obviously everybody's going to form an opinion on the partnership immediately uh even before they kind of understand the facts. But

as I looked at it I think it maybe was getting less hype than maybe it deserves because it's only a $1 billion investment. It's relatively small in the context of investment France.

But if you look at Disney's scale, the example I gave uh last there's like 140 million people have paid to visit a Disney park in the last year. I would imagine that all like I can imagine a world where over the next year every person that's visiting a Disney park is like hey I can kind of like personalize this experience and like take and basically create this entirely new kind of like entertainment layer to that. So I need a relationship with open one way or another.

Yeah. And I'm going to pay for the product. So it feels extremely significant in a way that maybe people aren't fully appreciating.

Yeah.

Yeah. I mean we're very very excited about it. If you think about like Sora and Image Gen, this is all about unleashing creativity. I think we're all born creators and like if you give people the tools to create easily and to go fast from imagination to a creation. We can see you know the golden age of creativity. But for that like you know we need people to have inspiration to have like you know like like IP to play with and Disney has by far the best IP by far the best inspiration and so partnering with them to bring these experiences to life is such a privilege and we're really excited about it.

Yeah. H how uh our audience I was thinking about this last night. Our audience has probably on average spent way more money on ads than actually seen ads because many folks in the audience have have spent money on on digital ad platforms, run ad businesses. Um how how do you think about positioning the ads product or whatever happens to the other side of the of the market like the brands that want to be successful in the future of commerce? Do you have any do you have any idea of of the the the shape that it will take because uh you know we were just doing a a Black Friday stream and and uh just one founder after another singing the praises of the the meta ecosystem even the apploven ecosystem. Uh the ability to go and and put money in a magical box and get customers is remarkable. uh have you thought more about what you want to offer ultimately to brands and companies that will be ultimately be partnering with OpenAI?

So not much to uh announce to the on ads but what I can tell you is two things. One like if we at some point we we uh decide to go towards ads we're going to do it in a way that is extremely respectful of the very special relationship that people have with Chad GPT. Like the thing to understand is people trust that charge has their back and gives them the best answer for their needs. Nothing we do can like jeopardize that. And so you know when I talk to brands and they're like hey how can we get more distribution? How can we uh you know influence the results of the LLM? The reality is have the best products like [laughter] you know we we want to build uh an LLM that is not impacted by uh you know paying companies that is not impacted um by you know people gaming the system and purely recommends the brands that are going to be best for you. Yeah, around that around that model response, you can imagine that there are things that we could do to get to to give brand distriution, but we really want to keep the model response pure because the trust that people have in that is critical.

Yeah. Is there is there a good like I don't know like how do you actually set up the company for that? You need like an editorial like firewall so that one team is just not even affected. They don't even like, you know, hang out at the happy hours together and say, "Hey, you know, can can you go do this this quarter because we have a big partner?" Like, have you thought about what what what it culturally takes to get there?

Well, the first the very good news at OpenAI is that the culture is already so focused on protecting the user, protecting the purity of LLM responses. It's so ingrained that I think anyone who would, you know, share like, "Hey, can you help me change things?" that's a that's a holiday party would be organ rejected. So I think that's that's already you know good. We built the right antibodies uh day one. The other thing is I think we're not just going to rely on culture. We're also going to rely on technical capabilities that have firewalls between these different things so that people can really trust that the model is not influenced by anything other than having their back.

Yeah. Yeah. That makes a ton of sense. Um, how are you thinking about uh the like setting up the actual applications team for success? Are you um I I I've been struck by the fact that the narrative has shifted in some ways from uh you know the best model wins every time. The best model, the best model, the best model, the best benchmarks like just win and you know and then everyone will just switch around. It feels like we're almost leaving that era where, you know, the the technical insiders, they do care about the benchmarks, they do care about the capabilities, but a lot of consumers just care about the user experience, the design, the the the the like almost the qualitative aspects. How have you thought about uh developing a like a delightful user experience? Is that a new muscle for the organization? Do you feel like OpenAI is set up to deliver, you know, this this delightful experience that can reduce churn and and keep people coming back? How do you think about that?

Well, first off, I would say, you know, this is a company that despite not starting as a product company has built the fastest growing consumer and enterprise business in history product with 800 million people. So, while I think I can, you know, build an organization that does that even better. I really want to celebrate what what they've done before me. Um, and and I think you're you're absolutely right. I think the magic is at the intersection of research and deployment. So, the thing I obsess about is how can we ti be tied at the hip with a product research team to understand what are the new capabilities that can unlock this new magical product. And you you kind of see that like sort of we were just talking about it a minute ago. It's a good example where the capability was there and and like that's what unlocks everything. But then the product thinking came in of like you know putting characters and yourself inside the video. And so it's really a mix of like capabilities but also a lot of delightful product thinking and really understanding what people want that unlocks these magical experiences. We were even talking at the beginning about kind of all of the use cases, right? Like

to me like you know just giving people more intelligence in a chatbot like that's going to cap out at some point. What I need to do is help them realize they can use chat GPT for health and like help them you know makes that easier, help them plan their travel, help them shop. Like all of these things are going to be based on incredible models but also great products that makes this understandable because right now you're still kind of having to stumble upon the value and like when people realize they can use strategy for health they're like oh my god this like unbelievable but they kind of had to stumble upon it and that means I haven't done my job fully until everyone knows how to use the value

you guys are are uh uh Sam was talking earlier this week with some folks about uh the push into enterprise. How how are you thinking about enterprise not on the API side but just on the on the uh core subscription size and around pricing uh historically as a you know 200 a month uh plan subscriber for Chat GBT. I had hit some points where I'd be frustrated because I could sense the model being uh lazy and I'm like I'm giving you as much money as you've you I'm on the highest plan. Can [snorts] I be on a plan that's just not never gets lazy?

This is the two this is the rumored $2,000 a month plan thinking like uh I just like can you make would it make sense to ever do an all you can eat all you can eat or consumption

based consumption on on on the non-API side just because you know I think consumption based is is wild. No one wants to do that but uh I don't know it's a bold bold statement. I I'd be interested to hear what you think. Yeah, you know, we've launched credit based pricing on the enterprise side so that in addition to your subscription, you can buy extra credit to get more. I think that's a model that could possibly work on the consumer side. So, you know, we we are open to exploring anything. We have it like now for Codex. Uh and so you know that's something that we could totally explore but I think it goes back to um a lot of the very I would say like sophisticated people really understand like the concept of buying more intelligence. Regular people just want like more outcomes. So I think the more we make that transition from a chatbot to like you give us your list of tasks and we do your tasks the more we're going to be able to make people realize the value and then you know the pricing will follow. uh your your original manifesto you outlined knowledge, health, creative expression, economic freedom, time, support. That that sort of covers everything. Is there a seventh capability that you're excited about? You mentioned spreadsheets. Maybe that's the seventh one right after time and creative expression. Put it at the top. That should be at the top. Economic freedom spreadsheets. Uh but I is there is there a seventh category or or a nent just even in your own personal use a new capability uh that you feel like okay you know maybe this year has been the year of agents we've gotten coding agents we've gotten deep research a year or two ago uh you know is there some new capability that you've seen glimmers of okay I think we're actually solving this piece or I think it might be good enough to use in this scenario. Yeah.

Yeah. I mean, we touched on a bunch, but what I would say is like for me the next step is unlocking connection to the ecosystem. So, like chatb is still fairly insular.

Yeah.

And we launched, you know, at dev day, we launched an app platform, but I think like we were talking about health, like health would be so much more powerful if you could connect LGBT to your health record and all of your kind of bowarables. Um you we were talking about like travel it's so much more powerful if Shept can connect with you know a lot of like uh travel companies and so figuring out how it can not just do things in this like you know text format but really connecting and doing things in the real world that's the thing that I think the models are ready for but we're still catching up on the product side to make that possible and unlock these really like you know much more sophisticated use cases. There's a ton of questions in the chat about the about the code red. We have to ask how is it going? Has it been an intense period for culturally? It feels like there's been, you know, a lot of ups and downs in any startup's history. Uh I imagine that it's been exciting and and and thrilling to to re-engage, but how has it been in internally so far? Well, for context, you know, code reds are not uncommon. Like, uh, we declare code reds as a way to really marshall a lot of resources towards a priority. It's a way to signal what's at the top of the list and what can be dep prioritized and really focus. And I think during this time, focus is critical. And in terms of how it's going, I mean, I'll let you be the judge of it, but I'm pretty proud of what we're releasing today. and like you know like leading uh leading the way in terms of you know best model for everyday professional work. That has been a goal of ours for a long time. It's been in the works for a long time. It didn't start with the code red but we're really excited to be able to release that today. And on top of that it's our 10 year anniversary today. So it's

wait today.

Yeah.

No way. I didn't realize it was today.

Get a trading card up ASAP team. What are we doing? 10 year anniversary of open air. That's incredible.

That's wild.

Fantastic.

That's [laughter] amazing. Congratulations.

How uh startups historically have uh been able to innovate because they tolerate some uh some amount of failure. And I think if you look at the hyperscalers too, they're willing to like launch products that uh that just don't work, right? If you look at the history of Google, they're they take moonshots. They make bets. Not everything works. Some do. And it's and it's uh uh an effective way to you know continue to innovate. How has I feel like when OpenAI releases a new product now whether it's Sora or the browser or anything like that it faces so much scrutiny from the world some people are excited and bullish other people want to see it fail. How how are you guys kind of what's your internal framework around failure today? Because ideally you want to keep shipping a lot of things and if you ship three things that don't work and one thing that does, it's like great, you have a new a new product that you can scale.

Yeah. Well, you know, I'm I'm no stranger to a lot of scrutiny throughout my career. I've worked at companies that that went through a lot of phases like that. Um, and I would tell you I think it's like it's simple and cliche, but you kind of have to block out the noise. The the thing that's good about open AI is like the northstar it's very is very clear right it's AGI that benefits all of humanity and we have a lot of conviction internally that we're on the right path to do that and that does mean that you know some things are going to work some things aren't and because the culture is routine in a research lab which is very different from other fear of failure is much less because people already know in research that you explore a lot of path and like many don't work like that's the core of research. And so because that's the DNA, people don't freak out when a product doesn't work because they're like, "Oh, it's kind of like research. We have to test a lot of things and some of them will hit, some of them won't." But but I think this innovative spirit has gotten us to where we're at and and really leading the market and I think that's something that we absolutely need to preserve.

Well, thank you.

Last question. Do you think are we still going to see age uh verification this year? Is that or or is that going to be a 2026 uh uh launch? So age verification is starting to roll out already in some countries and we're doing a slow roll out so that we can make sure that we're very accurate in predicting the teen with an adult and then the adult mode that's going to be unlocked by edge verification is going to land in Q1.

Great.

Got it.

Well, thank you so much for taking the time to come chat with us. Congratulations on all the amazing progress and such.

Congratulations to the whole team.

Congratulations on 10 years in the business of building AGI. We're feeling the AGI here. Uh, have a great rest of your day and we hope to talk to you soon.

Good chatting so much.

Goodbye.

Bye.

Let me tell you about Privy. Privy makes it easy to build on crypto rail, securely spin up white label wallets, sign transactions, and integrate onchain infrastructure all through one simple API. I love that. And I do the WWE voice