Cognition launches Windsurf 2.0 with an agent command center, bringing Devin into the IDE alongside local and cloud agent orchestration

Speaker 1: And Yeah. Great to finally meet.

Speaker 2: Great to have you on the show.

Speaker 1: Yeah. Super impressive.

Speaker 2: Meet. We'll talk to you soon.

Speaker 1: Cheers, Victor.

Speaker 2: Have a good rest of your day.

Speaker 1: Take care. Powerful nominative determinism.

Speaker 2: Victor. Yes. He will be the Victor. I like that. Well, our next guest is Theodor Marcu from Cognition. He's the head of product growth. Theodor, how are doing?

Speaker 14: Hey, guys. How's it going?

Speaker 2: It's good to It's see great. Obviously, everyone here on the show is familiar with Cognition, the makers of Devon. But take us through sort of how you're positioning the company, the updates, and any big announcements we should be aware of?

Speaker 14: Of course. Yeah. Thank you again for having me. Of course. This is a very exciting day. So, yesterday we had a huge moment for the company that I think a lot of us are very excited about. We had the biggest launch since the acquisition of Windsurf last This is something that the team has been looking forward to a lot. We launched Windsurf two point zero, which did two big things. It brought Devon to Windsurf, So, finally, the world's software is engineering available in Windsurf. And we have now an agent command center. So, our sort of vision for the future of software engineering, which is managing a team of agents, both remote and local, that works sort of alongside you have an army at your back. Windsurf two point zero makes that easier than ever.

Speaker 2: Got it. So, how what goes into an agent command center to make it effective? We were talking about Gastown and and having different agents for different tasks. The the I think the buzzword is like orchestrators. There's a various amount of, you know, open source projects and sort of the the idea of orchestration is percolating in the AI industry. How do you think about educating customers and enterprises about why they should be using an interface to manage agents instead of just having a bunch of different terminals open?

Speaker 14: Yeah. So first of all, this is something that I think frankly arose as a need internally at Cognition. What we're seeing is some of the best engineers here are, and we think this is sort of where the future is going, the best engineers are using local agents.

Speaker 2: Yeah.

Speaker 14: And they are planning their tasks, they're going very deep into the code base, architecting systems, coming up with a plan, and then they're handing those off to Cloud agents. And when we launched Cognition, Scott and Walden and Steven and few other folks had this sort of, they saw into the future and they saw that Cloud agents were going to be the future. But we also saw that engineers are sort of around and they will be, they are working very closely with their code and they want to make, there's sort of like golden age of engineering where we can of go very deep into what you're excited about and then hand out off to Cloud agents. So, with Windsurf two point zero, the bottom line is that, and what we're telling a lot of our customers with the command center, is that you get to have this overview of your agents so that you can, your limits on your attentions are no longer there. Because as you're working with dozens of agents at a time and you switch context, it gets really hard. So, we've built, we actually were very intentional and we built this Kanban view, sort of anyone that's familiar with project management can, has seen a Kanban view before, where you can see all your agents working on different projects at a time, and then you can sort of like switch in between them and quickly check on them and then spawn or create new sessions when you need to. Then those sessions are actually grouped into spaces, which is this way that we built to make it easier for agents to share context and also share sort of states together. So, from our perspective, what goes into a great command center is just making it very easy for a single engineer to work with a team of agents and have sort of like this army at their back.

Speaker 2: For for big enterprises, they have so much different development work to do. What is the the most low hanging fruit? Like, if you're talking to a customer who wants to get spun up with Windsurf and Cognition and Devon, are you trying to understand their their their backlog, the hairiest tasks, the most miserable? Are you trying to get them excited about greenfield projects and new dashboards and automated workflows? Are you trying to get in the hands of their their best engineers, their youngest engineers? Like, what is the greatest foothold for you right now?

Speaker 14: Yeah. That's a great question. I think that has That is something that has been evolving ever since the beginning of of Cognition. I think early on with Devon, what we we found was that we would go in and and talk about all these sort of very specific use cases Yeah. That we'd find inside companies, whether it's migrations, whether it's building internal tooling, whether it's these sort of like big backlog projects that they've been wanting to work on. I think more recently, as models have been getting better and as our agent harness has been getting better and better, what we're finding is that there's a lot of frankly, everyone can use a software engineering agent and the universe of possibilities has expanded greatly. And I was talking about how you can start working locally and, you know, there's this sort of like gold The way I think about it is there's this golden age of engineering where like the best engineers in the world can do more and they can offload their tasks. And not just the best engineers, frankly, everyone can offload the tasks that they don't want to spend as much time on to agents in the cloud that can handle them very quickly or over many minutes or long hours. And then they can sort of stay in control and work on the things that they care most about and the hardest problems that are most exciting to them. So, Scott always uses this sort of like our, idea of you're going from being a bricklayer to an architect. And I think a lot of what we're seeing a lot of our customers is that entire teams and individuals on those teams are moving from being bricklayers or sort of like writers to architects or directors where they're sort of like managing an orchestra of agents.

Speaker 2: Sure. Yeah. That makes sense. Jordy was asking this question earlier about just model agnosticism. How valuable like, do you see demand from customers and companies for wanting to switch between model providers, use open source for things? Like, how much is using the right tool for the job, understanding the Pareto frontier, and not blowing up your budget in one month if you're token maxing?

Speaker 14: Yeah. That's something that comes up all the time. Yeah. And as you know, Cognition has been model agnostic from day zero. Yeah. We've always worked with the best models and all of the models that are available, evaluated them internally, figured out how to make the best agent harness for Devon, how to make the best sort of agent harness for Windsurf as well. What we're seeing from our customers is that there's a lot of demand from trying out different models for different tasks. We're constantly working with them actually to make sure that we can advise them on what are the things that, know, models have this jagged line of intelligence. Some of them are really good at specific things, others are less good at those things. Like, for example, Opus 4.7 came out today and congratulations to Anthropic on a great model launch. Model is very good at deep investigation. So, we're sort of trying to look at how can we use that in our sort of, for example, Devin review workflows where Sure. Dev and can go in and look at a PR and try to figure out all the bugs and all the issues that might be associated with it.

Speaker 2: Yeah. Thanks. That makes sense. Well, congratulations on the progress.

Speaker 1: You Great so much for to meet you and give our best to the team.

Speaker 2: Yeah. We'll talk to you soon.

Speaker 14: Thank you, guys.

Speaker 2: Have a good rest of your day. There's a ton of breaking news. The big one is Reed Hastings. Reed Reed Hastings is

Speaker 1: stepping off the board of Netflix and the stock is down tremendously and but this is good.

Speaker 2: He's not stepping off the board. He's stepping off the board in June. He announced that he's stepping off

Speaker 6: the But yes, I mean, that's what

Speaker 1: I'm saying. No. But but but

Speaker 2: Why is

Speaker 1: this again? It's good because it's good for Reid specifically.

Speaker 2: Oh,

Speaker 1: yeah. Because it shows that people have confidence

Speaker 2: Oh, sure.

Speaker 1: Leadership and his vision. Yeah. And and and I'm sure that I would expect Netflix to make a quick recovery.

Speaker 4: But Yeah.

Speaker 2: It's up 12%, percent in the last month, down eight and a half percent overnight after hours. But we'll see where the stock settles, you know, tomorrow after the market processes this.

Speaker 1: The alternative is a nightmare for Reed because if he if he announced this Yeah. Announced this and the stock popped Yeah. 20% Yeah. He was, you know, handicap the company.

Speaker 2: Yeah. And I mean, flip side is that Ted Sarandos, it seems like he put on a master class over the last six months with the Paramount negotiations, not getting over his skis. The shareholders wound up really liking how that all penciled out. And so it seems like the company is in good hands. And all of the different strengths that Netflix has continue to show across advertising and subscriptions. And while they've kept their the big headline with Netflix is that they've kept their content budget essentially flat or slightly growing while they've grown subscriptions and revenue and top line very precipitously and very consistently even at a time where they've they haven't needed to invest exponentially more money in content. Obviously, spend a fortune on it, but it's not growing as fast as their revenue is growing, so their profits are growing, which is good news. Hastings departure marks the end of an era for Netflix, which under his leadership transformed from a d DVD by mail business to a juggernaut in subscription video streaming and disrupted Hollywood. He said, my real contribution at Netflix wasn't a single decision, Hastings said in a statement that was in a company letter to shareholders. It was a focus on member joy, building a culture that others could inherit and improve, and building a company that could be both beloved by members and wildly successful for generations to come. Well, we wish him the best on his next chapter, whatever he winds up doing. What an absolute run. Well, Jordy, is there anything else we should talk about? TSMC, of course, has earnings. We can cover those later. The chipmaker TSMC is more bullish than ever on AI despite the Iran war, so some good news there. And lots more stories to talk about, but we will be back with you on Monday at 11AM.

Speaker 1: It's been an honor and a privilege.

Speaker 2: And we'll see you soon.

Speaker 1: Be with you here today.

Speaker 2: Leave us five stars on

Speaker 1: Apple Podcasts. It's your day.

Speaker 2: Sign up for our newsletter at tbpn.com. Goodbye.

Speaker 8: Boeing flash bang.

Speaker 2: Boeing flash bang.