01:38:06 · Jeffrey Morgan, Jordi Hays, John Coogan
[01:38:06] S3: Yeah. I think you're right. It's oscillating. I mean, I'm a daily GLM five two user through Alomma right now and it's replaced 80% of my coding work. And I think a lot of that's gonna be true of a lot of customers. As for the gap, like to your point, I think it it may widen, it may shrink. I think overall it's shrinking. But but ultimately customers are gonna use a mix. And, you know, for the bulk of their use cases, they're gonna reach for these open models because they can tune them to be much faster. They're obviously much cheaper. And there's always gonna be use cases where you need the frontier. I don't if we'll ever settle the debate. I think ultimately the gap will will continue to shift. I think that's what makes it exciting. Right? It's like every every three months we're able to do something new. We're able to run better agents and quickly open models will catch up and really enable a whole wave of customers that wanna run open models to do that, you know, in their own environment or to customize it to the point where, like, they can even make it more more powerful. The last thing I'll say too is, you know, customers are readily taking these open models and customizing them, and they're actually getting better results often than [01:39:04] S0: Yeah. [01:39:04] S3: You know, just a a stock frontier model. Yeah. And, you know, I think we're just at the beginning of that transformation. [01:39:10] S1: Very cool. Well, congratulations on the progress [01:39:12] S0: Awesome to meet you. [01:39:13] S3: With the round. [01:39:13] S0: Congrats to
01:39:13 · Jordi Hays, John Coogan, Jeffrey Morgan, Thibault Sottiaux, Bernt Øivind Børnich
[01:39:13] S0: Congrats to [01:39:14] S1: the team. For coming on the show. And have a great rest of your week. We'll talk [01:39:16] S0: you Goodbye. [01:39:18] S3: Awesome. Thanks a lot. Cheers. [01:39:20] S1: Let me tell you about Shopify. Shopify is the commerce platform that grows with your business, lets you sell in seconds online, in store, on mobile, on social, on marketplaces, and now with AI agents. Jordy, there is a story. Oh, wait. We actually have our next guest. [01:39:34] S0: Guest of honor. [01:39:35] S1: Now, Tebow from OpenAI. He's the head of core products and platform. Tbo, how are you doing? Congratulations on the launch. [01:39:43] S4: Hey. Thanks. Doing well. [01:39:45] S1: Give us the highlights. [01:39:46] S0: Sleep last night. Do you sleep do you sleep at all? [01:39:51] S4: Yeah. I do sleep. Currently, we have, like, know, maybe five to 10 war rooms going. So, you know, it's just like a little bit it's intense. It's a sport. [01:39:59] S1: Yep. You [01:39:59] S5: know, but we're [01:40:00] S0: hurt to And and you guys like to make it hard on yourselves by launching every every time there's a launch day. It's like, you know, 15 new things. So it makes sense that there's, you know, close to equal amount of war rooms.
01:40:00 · Jordi Hays, John Coogan, Thibault Sottiaux
[01:40:00] S0: hurt to And and you guys like to make it hard on yourselves by launching every every time there's a launch day. It's like, you know, 15 new things. So it makes sense that there's, you know, close to equal amount of war rooms. [01:40:12] S1: Yeah. Let's start with 5.6 Soul, though. I want to I wanna I want you to identify for me, like, what is sticking out? What are the most cutting edge capabilities that really stuck out to you? The latest unlocks from the frontier of the actual model. Then we can go into codex and voice model and everything else and how things come together and how these are used. But first, just from the raw model capability, what was most impressive to you? What was most exciting? [01:40:41] S4: Yeah. It's it's it's actually really hard to answer that question, Krishna, because when we were looking at the benchmarks, trying it, you know, we were just, like, blown away by it all. It's just better at coding, better at cyber, better at everything like long context, producing documents, better at having taste, website generation. And then for the first time, we also really cracked, I think, multi agent setups, which we shipped the ultra mode. And when you just see that going and, you know, you've got like eight agents collaborating together, communicating, and, you know, getting the same work done like faster, it's just you just feel like, wow, you know, this is like another way to scale test time compute. [01:41:17] S1: Yep.
01:41:17 · John Coogan, Thibault Sottiaux
[01:41:17] S1: Yep. [01:41:17] S4: But overall, just amazing workhorse feels way, way better than five five and anything else we've produced so far. [01:41:23] S1: So, in January, there was a project that was getting some attention called Gastown, talking about all these different sub agents. You had pull cats and all sorts of different abstractions. It felt, highly technical, and it seems like Soul Ultra is a way to abstract that away. Is that is that deliberate? Is is there, I guess the question more broadly is, like, is the level of, like, prompt engineering, are we are we leaving that era, or or will there always be some cycle of, you know, if you get really good at using Soul Ultra, you'll have a better experience because you'll be able to give more fine tuned, fine grained prompt and and direction to the model.
01:42:07 · Thibault Sottiaux, Bernt Øivind Børnich
[01:42:07] S4: So one thing that you see as well with Soul is its uncanny ability at understanding human intent. And you know, you need shorter prompts, you don't need to explain yourself in that much detail and sort of like, you know, just gets it and then goes and like does like a very complex thing. You know, you saw the prompt for like post training, the Luna model, which is like super crisp, and then it does that, but it actually worked for many days. Yeah. And this is also like with us launching Chat50 Work, it's about making it accessible for everyone and you know, you don't need to have like a PhD to use this model. It's just like, should just behave like just like another super super smart human and you know, just kinda like get you in the moment, and that's what we're striving for. Of course, if you know, you really push it to the limit, you're always going to find new setups and this is like also a very exciting space. Know, we continue to develop like also codecs in the open source and we're seeing like, you know, all of novel ways to set up these agents and models so that, you know, you can get results in like cybersecurity and like, you know, all these other more nuanced and complex things. But for everyone, you know, you should just feel like, you know, a ton of power out of the box. [01:43:16] S5: Mhmm.
01:43:16 · Bernt Øivind Børnich, Jordi Hays, Thibault Sottiaux
[01:43:16] S5: Mhmm. [01:43:16] S0: Talk about what was important at the product layer. Fundamentally, what I think people want out of products is to just be able to talk to their computer like a really smart coworker and be able to get things done. But then you're dealing with, you know, so many users over here, millions of users over here trying to combine it and condense it into something that's simple and obviously simple things end up being exceptionally complex to actually create. [01:43:43] S4: Yeah. So if if you look at it, it's deceptively simple. You can open it on your phone. It's ChatGrifty work, you just toggle it and then there you go, you connect it to the things that you already have, your email, calendar, you know, your docs, and then suddenly you're like, okay, wait, I can ask it to process all of this information that I had over there that I had to manually like do all these things myself and it can just do all of that. And it's just like on the go and it's like on your phone in the Chateapute app you already have installed. I mean, that's the beauty of keeping it very simple. At the end of the day, we want it to be just a normal conversation between you and the agent. This is also why we decided to, you know, ship it, like, just in touch with you.