Agent orchestration is so two-years ago

**Ryan Donovan** (0:00)
Solving difficult problems is what drives you. Raytheon is hiring software, systems, and hardware engineers to take on some of today's most complex technical challenges. Explore opportunities at careers.rtx.com/raytheon.
Hello, everyone, and welcome to The Stack Overflow Podcast, a place to talk all things software and technology. I am your host, Ryan Donovan, and today we are talking a little hot take that people are developing agents based on 2024 models and not the modern models. Making a big mistake, everybody. So my guest for that is Saahil Jain, who is the CTO of u.com. So welcome to the show.

**Saahil Jain** (0:48)
Hey, thanks, Ryan. Really excited to be here and chat with you about all sorts of topics.

**Ryan Donovan** (0:52)
Yeah. Well, we got a bunch of stuff to talk about, but before we get into it, let's get to know you a little bit. How did you get into software technology?

**Saahil Jain** (1:00)
Yeah. So it's a little bit about me. So I'm currently the CTO of u.com, an AI search company. My background is a bit windy in some ways. I initially, I kind of start my journey. I was doing research in AI at Stanford, where I was focused on AI and healthcare. And while doing this, I ended up focusing a lot on natural language processing. So a lot of my work and research focused around developing really good ways of extracting information from radiology reports. And I developed a lot of models. This was in Andrew Ng's Stanford Machine Learning Group. And it was a lot of fun. And as I started diving into natural language processing, this is around 2019, 2020, the transformer model started to show really good results. And you can see the rise of the BERT-based architecture.
And this is when I got acquainted with Richard.
So Richard was kind of the co-founder and the CEO of u.com. And I ended up deciding to focus on what I thought was the most interesting application of natural language processing, which is search. It brought me to u.com, where I've been focusing on search and AI ever since. So for the last five to six years, now I think my journey in software and technology is really centered around focusing on topics related to information retrieval, search, and AI.
And of course, it's changed a lot over the last five to six years as the industry has grown.

**Ryan Donovan** (2:15)
Everybody's jobs have changed a lot in the last five or six years.
So you talk about focusing on search. Obviously, search has become a pretty big use case for AI since it has a sort of natural language understanding. What is it like building an AI-focused search?

**Saahil Jain** (2:33)
So search has definitely evolved a lot. We actually started off as a consumer search engine back in 2020
And that was when the main user insight was a human.
And now, fast forward to 2026, and the main users of search we believe will be AI agents. Fundamental paradigm shift. So what it means to develop, I think what you described as AI search, is search geared towards the new consumer of the Internet, which will be AI agents. We expect that most people will be interacting with AI directly. And this AI will, in some ways, act as an intermediary between the various tools that we humans have spent so much of our time navigating around, like search. So a big part of AI search is developing search for this new consumer, which is AI. Yeah.

**Ryan Donovan** (3:25)
That's an interesting way to structure search. I assume that the thing that you're giving AI search is not just like a massive rag system, right?

**Saahil Jain** (3:36)
I think when it comes to search, if I kind of back up a little bit in terms of how I view the whole space, is I think that at very high level, we think of there as being two really big types of horizontal AI initiatives, one of which is what we can call core intelligence. These are the foundation models that the companies that are building these very intelligent systems, open AI, entropic, in mind, focused on building out these foundation models that are very intelligent. And then on the other side, we have what I would call augmented intelligence companies, companies that are focused on building the tools and services that guide some of these foundation models. What we're focused on is really building out the tools and services that can guide core intelligence. This is a very interesting space because in many ways, the way I like to think about the industry as a whole, is it's almost the reverse of self-driving. So when it came to self-driving cars, we had the roads for a while, for many years, and then we've recently developed these autonomous systems that can navigate the roads.