Head of ChatGPT & Codex: agents for normal people are HERE — Silicon Valley Girl Podcast

Marina Mogilko: There is going to be dramatic change. Everyone is going to get their own little personal assistant on their computer that allows you to do more than whatever was possible 3 months ago, 6 months ago.

Thibault Sottiaux: This is the head of ChatGPT, Codex, and the APIs at OpenAI, the tech powering most of the AI you already use every day. And he just told me that in a few months, people who don't use AI at all will get the same benefits as the ones who've spent the last 2 years figuring it out.

Marina Mogilko: Today, you have to actively prompt and be creative about what you're asking. In the future, that's not the case anymore.

Thibault Sottiaux: What's going to make people brilliant? If you're a creative person with good ideas, it's a great time to experiment.

Marina Mogilko: Can you show me a workflow everyone should be deploying today?

Thibault Sottiaux: One thing that I think this will mean is that my audience is some of them are developers, but a lot of them are knowledge workers and they're very excited about AI. And what we've been seeing with software engineering, Google says 75% of their code is now AI written. So it has been really transformed by AI and you said something that this transformation will extend to all knowledge work over the next 6 months. So what do you think is going to be changing for people who are knowledge workers? I think it's not that much that people are going to change as the technology has matured and now agents are reliable over long horizons of capable of using many different tools, computer use being one of them, browser use we have added over 100 different plugins that can tap into every little tool that you already use in your life and the agent and GPT-55 is extremely reliable at it so it's more that the technology has matured and is ready and so everyone will be able to get benefits from an agent. Whereas before you had to be a little bit technical and get in there if it wasn't reliable if it was struggling 5 minutes in you maybe have to go into the configuration and obviously that requires a technical background that's not the case anymore. So that's why I think it's going to become very widespread.

Marina Mogilko: What do you think is going to change the most and say if you take a marketer what's going to change their day-to-day with agents?

Thibault Sottiaux: So you're going to look at what can you automate? Maybe I spend 1 hour doing market research. Maybe I spend one hour summarizing a whole bunch of inbound emails. Maybe I spend two hours going through prospects. And so a lot of the concepts that were advanced concepts maybe six months ago, such as running something on a cron schedule, nowadays it's just in the app, you can just say run this every 12 hours, do some market research, send me a PDF.

Marina Mogilko: And it's going to send it to you over email.

Thibault Sottiaux: Yes, send it to you over email or you can consume it in the app maybe you can print it. I did this a couple of weeks ago where I would summarize all the news from Slack and then print it on my printer physically every day so I would have it with the coffee.

Marina Mogilko: Oh wow. This is very executive, just wanting my newspaper, this kind of thing is going to become mainstream.

Thibault Sottiaux: You know what I was really inspired by? I was watching one of your podcasts where you were talking with Greg about having this dashboard basically when you wake up in the morning, you have your coffee and it just tells you can you approve this and I'm going to run it. Can you approve? So you just read through whatever your AI agent did while you were sleeping and you're approving and it's continuing its work. When do you think it's going to be a reality?

Marina Mogilko: It's not that far off. The technology already exists. It's about packaging it.

Thibault Sottiaux: For example, last week we released auto review which is a concept that definitely blew my mind when we did the research on it. So you have the main agent doing actions and then you have a second agent that is verifying all the actions of the first agent and verifying that they are not doing anything that could be harmful to you and are low risk. And so that's an innovation that came out of our safety team and alignment team primarily alignment research and it allows you to run an agent much more for longer periods of time autonomously even handling quite sensitive data without risking for example your agent sending an email to a stranger with some of your personal information in there. Now that you have these pieces to make it more secure, more safe and you're willing to give access to a lot of things in your life.

Marina Mogilko: How do we organize our data? Imagine we're all getting ready to deploy a first agent. The thing is, an agent needs good data. Sometimes I've been working with mostly my tone of voice like personal dossier, but then I realized that my agent would really benefit from knowing my strategy when it creates content for me. So what would you say to someone who's just starting out? How do they organize this data? What are the must-have files that they have to create?

Thibault Sottiaux: I organize everything in local files on my computer. So I have a folder with my notes in there. I keep it fairly tidy and I use my agent to organize it and it grows over time.

Marina Mogilko: That's my question. People work with multiple computers. I have a laptop that I travel with. I have two Macs, one in my studio, one in my office.

Thibault Sottiaux: Yeah. So what we're going to see over the next three months is all of this is going to come to the cloud and you're not going to have to manage local files on just your laptop and then if you happen to travel and you're on your phone, now suddenly it's not the same agent. And that's annoying because you have two separate entities that you need to map to different things in your life. So a lot of this is going to come to the cloud. It's going to manage its own memory, manage help you manage files that are just hosted somewhere.

Marina Mogilko: There's no good solution yet, which is surprising.

Thibault Sottiaux: Yeah, I'm just putting everything into my Google Drive and I'm connecting a Google Drive folder.

Marina Mogilko: Yeah, that works. What are the files that everyone should come up with for their agents?

Thibault Sottiaux: If we're talking about productivity for knowledge workers, tone of voice is a good one. And for that, surprisingly, I don't recommend trying to explain your tone of voice. I recommend including examples of past newsletters that you published or snippets from recordings or messages that you sent to friends in different settings, maybe professional and personal. I have a lot of notes around projects that I have, various each project has its own folder with various files in it.

Marina Mogilko: Yeah.

Thibault Sottiaux: And then contacts but you don't have to maintain everything in files. You can also rely on all the productivity apps that you already have and then Codex for example will be able to go and pull the right information.

Marina Mogilko: How do you know when you stay inside a project or you want Codex to actually build something like a tool or an app? Sometimes I feel like I need a dedicated app for a new use case and then I create a new folder and just experiment in there. But nowadays more and more I feel like I need less of static apps that just can do one thing and I rely just more on my agent to do everything for me.

Thibault Sottiaux: If somebody starts using agents and somebody doesn't, what's the difference in their productivity in one or three years? One to three years is so far off. The technology itself will change so much by then. What we're going to see is people who are willing to adapt and discover these things are going to be quite a bit more productive and it's going to enable them to do all these things that they maybe put off. Everyone is going to get their own little personal assistant on their computer doing all of maybe filing your taxes setting up email filters telling you and helping you get more in touch with your loved ones. That is something that will connect a lot of people.

Marina Mogilko: When I started using more AI, I felt this responsibility dilemma. You just mentioned taxes.

Thibault Sottiaux: Right.

Marina Mogilko: AI can totally do taxes for me and I have this automation that tells me every month how much tax I'm going to owe and what tax strategies I could be deploying. Do I want to be responsible for that? I don't know. When it comes to these tasks, also as an entrepreneur, I totally realize if you're just starting out today, you're much more productive. But if you're an existing entrepreneur, I wouldn't really fire anyone on my team because an AI agent can run it. I would just ask that person to use that AI agent. Where do you see this go? Because I feel like when I started running so many agents, I feel responsible that I need to check on them. I need to verify their output. I'm like, oh, I'll just let somebody else handle it.

Thibault Sottiaux: That's interesting. So at the end of the day, humans remain responsible. And the way that we think about it is actually just really about augmenting your own capabilities and giving you a tool that allows you to do more than whatever was possible three months ago, six months ago and take a lot of the boring parts of the work and just do that automatically for you. It is important as we see this for coding—if you produce a piece of code you know you are responsible for it. If it breaks it's not the agent's fault, it's your fault. Same for code review and you cannot outsource understanding. At the end of the day humans remain in control and remain the people who need to understand the entire system and how things work. So if you view it under the lens of augmenting your own productivity as a human, you need humans after all. We're building for ourselves to better our own lives.

Marina Mogilko: Yeah, it's just I feel like I become much more productive first, but then because I want to optimize so many things, my brain just started exploding.

Thibault Sottiaux: There is a pitfall of falling too much into using it for everything and feeling like maybe you can do things but you're a little bit too early on the capability curve. So maybe things will be possible with future models. That's good. That means that you're pushing and discovering where it's reliable, where maybe it needs a little bit more work and maybe this will be something that you can do in 3 months, 6 months.

Marina Mogilko: Today's conversation is another reminder of how much AI has sped everything up. My calendar this year has more meetings in it than ever before. So I've been testing a lot of different things to keep up. A couple I've been using lately, the Soundcore Liberty Pro Series, Liberty 5 Pro, and Liberty 5 Pro Max. Both have the same noise cancellation and call quality. The Pro Max is also equipped with built-in AI functions designed specifically for business people, and several other features have pleasantly surprised me. First, call quality. I travel a lot and often take calls and attend meetings on the go. I take calls from planes, airports, moving cars, and around kids. With regular earbuds, people often couldn't hear me clearly. With these, the microphone picks up my voice and cuts everything else out. The earbuds hold a Guinness World Record for call clarity in noisy environments. They were tested against 13 competitors by an independent lab and scored first. Second, the AI Notetaker in the case. Tap and it starts recording. After the meeting, I get a summary, speaker breakdown, and a to-do list straight to my phone. The whole thing lands there automatically. And what else caught me off guard? The noise cancellation. I put these on during a long flight and genuinely forgot there were hundreds of people around me. Their app has built-in sleep sounds and meditations too. I sleep a lot on airplanes and hotels, and this is a lifesaver for me. You close your eyes and you actually get to properly rest. If you're running a busy lifestyle with back-to-back calls, no time to take notes, Liberty 5 Pro series are worth trying. 30 days free trial, easy returns if you're not satisfied. The link is in the description. Now, back to our conversation.

Marina Mogilko: Okay, let's talk about no-code in general. I shared before the podcast that in Lingu, we're actually no-code a lot of different apps and we have a technical person who can actually write code and he's really good at it, but then it's also us who can't. And when we no-code something, if it works great but then the technical person comes in and says oh I would have structured it in a different way. Because if we wanted to grow it—for example we have this app that is 300 most useful words in English and the way it was no-coded in a perfect way so it's usable but if we want to extend it and make it like a thousand useful words then the architecture is a little bit off. What would you say, how would you approach no-code these days? If you're thinking of building something like it's going to be a big company, should you work with a technical person or still try and no-code it?

Thibault Sottiaux: If the goal is to share it with a couple of people and try something out and build it for the joy of it, definitely cutting it by yourself with the help of an agent. That's perfectly fine. If the goal is to scale it and make something that is successful for hundreds of thousands of people or more, having a technical person in the mix is still useful. I do expect that over time to go away where you will have agents which are able to understand the long-term maintenance aspects, the right structure and be able to handhold you all the way to scaling to very large and successful products.

Marina Mogilko: What do you think is the timeline because it's been changing so fast.

Thibault Sottiaux: We will see significant improvements towards long-term maintainability of code over the next 6 to 9 months in order to not need the help of someone more technical at all. That feels quite far off.

Marina Mogilko: What do you think in general about the future of software engineering? I was just moderating a panel around it with deep learning AI.

Thibault Sottiaux: And we're talking about how we all had a consensus of six people on the panel that you still need technical knowledge but you also need to become a generalist. Because everyone is no-coding apps, as a creator I would have never thought of building an app 10 years ago because I knew how expensive it was. Now I would build an app but then I will see thousands of people using it and I will experience some bugs and I will need a technical person. Are we going to need more software engineers or less with this?

Marina Mogilko: Yeah, I think we're going to see an explosion in the amount of infrastructure and apps that we built. Nowadays it's awesome because if you just have an idea, you can build it or at least in some shape you can build it. If you're a creative person with good ideas, it's a great time to experiment and then you can quickly prototype, iterate and get to something that you feel might be successful out there. Then it's hard to put a limit on how much software is enough. How much demand is there for more software in the world? How many more problems can we solve? It feels like we're at the very beginning of it.

Thibault Sottiaux: Feels like we keep coming up with problems we can solve.

Marina Mogilko: Exactly. So as long as that continues to go through and we are able to continue technological advancement, it feels like there will be demand for a lot of technical folks.

Thibault Sottiaux: Can you show me a workflow that everyone should be deploying today to maximize their productivity? Can you think of something?

Marina Mogilko: It really varies. One of the things that I do is I use it to stay in tune with what the world thinks, for example about ChatGPT, and so I receive a daily summary of something for me.

Thibault Sottiaux: So let's do a new chat for example. I am on a podcast talking about AI and CodeEx. Find relevant emails in my inbox and prepare draft replies based on what you know about my priorities from memory. So you haven't built memories yet.

Marina Mogilko: But if you use CodeEx a lot it will learn about memory from ChatGPT as well.

Thibault Sottiaux: Memory from ChatGPT as well, right.

Marina Mogilko: Yeah. Another one is go through internet sources and give me a breakdown of what CodeEx shipped over the last two weeks. And we can go back to email. I have a messy inbox in Gmail. Set up filters for it to keep things organized.

Thibault Sottiaux: I remember I think it was like 2019 or something. I spent a couple days just organizing my emails and folders and then I realized it's such a waste of time because I spent so much time on that.

Marina Mogilko: To plan a trip, where should we go based on my calendar availability.

Thibault Sottiaux: Based on your calendar availability, right?

Marina Mogilko: Or something.

Thibault Sottiaux: Oh yeah. For example, it can connect to my calendar, a trip plan, and use my calendar availability to plan it. For example, so it's going to go like here—you can see it's gone through your email. So this one is like I'm a podcast talking about AI and CodeEx. It's already gone through your email, found several live threads related to today's thing.

Marina Mogilko: Head of ChatGPT & Codex: agents for normal people are HERE

Thibault Sottiaux: It went through the calendar. It's all still going through emails. So it found a bunch of things. Apparently you have Codex which is today, and then here you have another one. So you can see we're doing a lot of things in parallel. You can just run little agentic threads. Here you could say I am making a presentation, put this in Google Slides and it will just prepare a Google Slides for all of the content.

Marina Mogilko: The editing is going to be within the app or I will have to leave and go to Google Slides.

Thibault Sottiaux: So there is an in-app browser as well.

Marina Mogilko: I don't know if you've set up your pet. No, not yet. So we also shipped this cute little feature. I just saw it on your X. So now you can move this across and you can see which threads it's still working on. For example, it's still summarizing Codex updates. It's currently setting up some email filters.

Thibault Sottiaux: I don't want to make changes to your inbox, so I'm just going to stop this one.

Marina Mogilko: Can you talk to me about this dashboard that I was looking at?

Thibault Sottiaux: You have a trip to Creed's plant now.

Marina Mogilko: Nice. So it went through my calendar. So this morning I actually asked ChatGPT based on my queries what app should I build and it gave me five ideas and this was one of the ideas where it's content repurposing.

Thibault Sottiaux: That's cool. I created this app, but the only thing I wanted to be able to dictate into the app so it can convert it into an actual email newsletter, but I don't think it worked.

Marina Mogilko: So let's have a look. So this was content repurposing engine from all the posts. And then it just built all of that for you.

Thibault Sottiaux: Well, it built something too fancy. I'm like, let's just make it very simple.

Marina Mogilko: You were like, okay, this is a little bit too ambitious here. AB testing.

Thibault Sottiaux: It's already a newsletter draft.

Marina Mogilko: Yeah. It actually I like that it didn't wait for my prompt. It just did a lot of these things by itself. It was this is probably what you need.

Thibault Sottiaux: This is cool. The only thing that doesn't work is being able to dictate into the app.

Marina Mogilko: Is it something that another prompt is going to fix or do you want to speak to the voice dictation here?

Thibault Sottiaux: Yes. Let's use it. Let's do a nested one.

Marina Mogilko: Okay. So in the app we don't have dictation working. Can you integrate it with the OpenAI speech-to-text API? For that you will need an API key, but you can just make it up for now. And you can go through the latest documentation online to figure out how to integrate the OpenAI speech-to-text API into this app.

Thibault Sottiaux: That was a very sophisticated prompt. How do I know that I needed that prompt? Because I would never think about API keys.

Marina Mogilko: Oh, that's why sometimes you still need someone who's a little bit technical. But a lot of these things that right now maybe are still a little bit in the realm of oh you need more of a technical background are going to become mainstream. So maybe you need me today but in three months it will be able to set all of this up. It will just connect it to your OpenAI account and you won't need to set up a separate API key.

Thibault Sottiaux: And it's going to be able to just connect to everything. While it's building, can you talk about the most sophisticated use cases or maybe use cases that are very sophisticated in terms of workflow but that change productivity tremendously for people day-to-day, something that everyone could get inspired by?

Marina Mogilko: So you can say go through and you can tag a lot of different plugins. So you can say Gmail, calendar if you have docs, and be my chief of staff and give me a breakdown of my day, a summary of what is important and prepare me. And then you just run this at the start of the day and then you can be more specific than the example prompt that I gave here. Some people that I've seen have fairly sophisticated requests for how to be prepared.

Thibault Sottiaux: And like where am I wasting my time? Should I be hiring someone to fix something where I see I'm spending most of my hours or building an app that's going to optimize it?

Marina Mogilko: You can really discuss with Codex then and start to brainstorm about how you can streamline things.

Thibault Sottiaux: Your calendar says this part has high consequence. So it seems like you're doing good. I only put important stuff in the calendar. I don't put day-to-day.

Marina Mogilko: What do you personally use and in what order for which tasks?

Thibault Sottiaux: So I use it for coding a lot. I use it to help strategize for the different pieces that we're planning to put out, what order, the narrative to construct. So it's a great thought partner. I use it to organize all my own notes throughout the day. So instead of taking notes in a separate notes app now I just take them straight into Codex and it just builds up memories.

Marina Mogilko: You just start a new chat for anything you want to say.

Thibault Sottiaux: I'm just like hey this is what I'm thinking.

Marina Mogilko: And then at the end of the day does it automatically summarize it or do you have to prompt it?

Thibault Sottiaux: I prompt it to dump everything into a doc. But also it builds up memories. I use it to often stay on top of unanswered important emails or Slack messages.

Marina Mogilko: Are you seeing the shift from more code work being done with AI to more productivity stuff or has it been equal?

Thibault Sottiaux: So it is interesting to think about what does an engineer, what does a programmer do all day and the majority of the time is not spent on coding actually. We saw a shift over time, even in our technical user base, over 50% of the tasks done on Codex are non-technical tasks today.

Marina Mogilko: Already.

Thibault Sottiaux: Yes.

Marina Mogilko: And it's been what, a year since you launched?

Thibault Sottiaux: The app has only been three months.

Marina Mogilko: Okay, we launched around three months. Codex was definitely for programmers, that was a year ago. So what are people using it for, what do you see, what are the use cases?

Thibault Sottiaux: There's a lot of help me organize thinking through things. Now with computer use and browser use, it can do things such as order on DoorDash for you, do your shopping.

Marina Mogilko: Does it take long? I haven't tried it on browser yet, but sometimes the agentic browsers are slow.

Thibault Sottiaux: No, we have the fastest computer use implementation out there.

Marina Mogilko: It's surprisingly quick. So you will see it click around. For example, I can ask it to go and download my LinkedIn analytics.

Thibault Sottiaux: Okay, so let's do add computer. And then go to LinkedIn and download my analytics in a spreadsheet format.

Marina Mogilko: Create a spreadsheet. There you go. So let's see what it does. I'll use the desktop controls. This might take five to ten minutes.

Thibault Sottiaux: So the difference from ChatGPT is that it can use your computer. Yes, it can use your computer. It's a full agent that works and has access to your browser and can click around on your apps as well. You can see it here. It's already visiting.

Marina Mogilko: It's already on my LinkedIn page.

Thibault Sottiaux: Yes. So let's see what it does over time. You can see it's navigating here.

Marina Mogilko: Exporting something.

Thibault Sottiaux: It is already saving stuff. That's amazing. While we wait for it to finish, I wanted to ask you something. You dropped out of your PhD, right? After two weeks?

Marina Mogilko: After two weeks, that's right. I started a company. Would you recommend to someone who's doing their PhD or bachelor's, master's now and feels like they're just hating it? Is it a good time to quit?

Thibault Sottiaux: This was a long time ago. This was almost 15 years ago. So I'm not sure.

Marina Mogilko: It's interesting how the meaning of education has changed in these 15 years. What do you think from seeing all the innovation inside OpenAI, Deep Mind? Should people be still pursuing very deep education or should they just go out and start building?

Thibault Sottiaux: The reason I dropped out of my PhD is I felt that committing four years of my life to that specific subject was not what I wanted to do.

Marina Mogilko: And I had more ideas and more things that I wanted to try, and so I felt the startup environment was right for me. I think it's very important, time and time again in my own life, I've realized it's important to follow your instincts.

Thibault Sottiaux: And do things that give you energy.

Marina Mogilko: Was it the right decision for you?

Thibault Sottiaux: Oh absolutely. I had such a great time. I built that company for a year and then I went and worked for Google. Initially I was working on Google Maps and different parts of ads, and then I ended up working at DeepMind, which was during the golden ages of DeepMind, when it was grand challenge after grand challenge, and then it became a little bit less interesting. Around the time when everyone was scaling and OpenAI had proven that you could scale transformers and create something very compelling, and this was probably the road to building AGI, that's when I was like, okay, let me go work for OpenAI. And time and time again, I challenged myself to pursue new things.

Marina Mogilko: So before we look at what has created, one final question: what do you think is going to happen to our lives in 3 to 5 years from all the innovation that you've seen? Do you think we're going to see this dramatic change that a lot of people are talking about, or are they exaggerating?

Thibault Sottiaux: I think there is going to be dramatic change. I think it's important for us and for me personally to really bring the benefits to everyone, and one thing that I think this will mean is that you will have the benefits whether you're actively engaging with this technology or not. Today you have to actively prompt and know how to be creative about what you're asking, and the benefits that you get are proportional to your prompting. In the future that will not be true. It will be very similar to going to a nice tailor who instantly looks at you and gets you and says, "This is the kind of clothes that would make you shine, and this is just right for you."

Marina Mogilko: So what will it depend on then? If it's not your prompting, what's going to make people brilliant?

Thibault Sottiaux: It will be much more about benefiting from good advice from a friend or having a natural conversation like the conversation that we have right now.

Marina Mogilko: So the ability to ask the right questions, not to necessarily prompt them.

Thibault Sottiaux: It will be about engaging as your authentic self in a conversation and then getting the appropriate help at the right time.

Marina Mogilko: And benefiting from this ambient intelligence that is just supporting society. Okay, it has created a spreadsheet. So it has downloaded all the analytics from...

Thibault Sottiaux: Oh, that's amazing. Well, it's okay. We needed this. Like, if I ask it to do something more specific, because it did one screenshot, but we can.

Marina Mogilko: What else would you like it to do? We could run it again. So I need more data, for example.

Thibault Sottiaux: Impressions per post. Okay, so I am very lazy now with Codex and I just do everything by voice. So I need more data: impressions per post.

Marina Mogilko: Same. I just use Whisper flow across all my content. Okay, let us end this daily brief. It did it. This is getting better.

Thibault Sottiaux: And then you can continue and get more advanced if there's a workflow that you like. You can use the skill creator, which is a skill by itself, and then you can say, "Create a skill to capture this workflow such that I can run it every day," and then it will create a bespoke skill for you that you can run.

Marina Mogilko: Collect analytics and add it to file.

Thibault Sottiaux: Okay, this is fascinating. Thank you so much. I feel like I have some work to do when I come back home tonight. This is really fascinating. Thank you for building it.

Marina Mogilko: Of course. It was a pleasure. I also write a newsletter where I go deeper on AI tools that I use, career strategies, and things I can't fit into a 30 minute podcast. It's free. Link is in the description.