This conversation spotlights Jeff Wang, CEO of Windsurf, who provides a behind-the-scenes account of navigating Windsurf’s reverse-acquihire by Google and merger with Cognition. Jeff’s human-first, pragmatic leadership in the face of adversity shows why every startup needs a Jeff!
Transcript
Jeff: I just didn't have time to panic. I think that was all like the only thing in my head was just like, be relentlessly. Focus on just finding a solution.
Selina: Welcome to On The Fly, a conversation series with leaders about critical decisions they're making that shape the future of their businesses. I am your host Selena Wang, and I'm joined by my co-host Palak Goel. Today we have our very first guest. Jeff Wang, the CEO of Windsurf. We chose Jeff to be our first guest because he recently made a series of critical decisions that's impacting the future of Windsurf in a very unique way.
This conversation is a deep dive into his leadership and the future of windsurf in coding ai. Jeff, we are so thrilled to have you here. Welcome to the show.
Jeff: Thanks for having me.
Selina: So Jeff, you've been talk of the town for the past month, however, people haven't really had a chance to get to know you. So why don't we start there?
Where are you from and what brought you to the Bay Area?
Jeff: Yeah, I mean, I'm from the Midwest, uh, grew up in Chicago. Mm-hmm. Uh, you know, as a kid I was always trying to build things and like make movies and shoot videos on like, I think my dad had like a camcorder that was like an eight millimeter. Tape and I was trying to upload it into the computer, and it was only two gigabytes of like, hard drive space.
I remember this like, it's also dating me probably. Um, but like that, that's like, kind of like how I grew up. And he said, you could be anything you want. Um, as long as it's, it's an engineer. Um, so I, I did go to school in Michigan for engineering and then, uh, got my first job into the Bay Area and I've just been here ever since.
Um, and I've just. Held jobs at like big companies first and realized it, it was too big. Uh, yeah, like Cisco and Salesforce, and then held several like executive positions at like startups. Uh, and then finally kind of found windsurf, uh, then it was podium, uh, and uh, I think, I believe that was the first podium hire actually when it went ahead.
Its first pivot. Yep.
Selina: You joined Codeium back in 2023, what brought you to this company?
Jeff: You know what's crazy? Um, I, I knew about GPT two, uh, when it came out from OpenAI. Mm-hmm. I was, at the time, I think it was like 1.5 billion parameter. Again, like at the time it was like, wow, this is, this is a really cool model.
Um, and prior to that, when I was at Salesforce, I was trying to like figure out how do you. Make models for different like classifiers, like for example, like a, a support ticket. What kind of a topic is this? Or even like positive or negative sentiment. Um, these things were like, you, you would dedicate tons of resources to, to create these models.
And then with, when GPT two came out, it was like, wait, you can just use one model for. Everything. This is insane actually. Um, and then actually like, um, as GBT three came out, uh, the llama models started coming out and I was just at home like tinkering with llama one. Mm-hmm. Uh, I had a pirate, I think it was only on BitTorrent or something.
Um, so I was playing with like the llama models. I was playing with all the diffusion models for images. And I just knew like, this is definitely the future. I just had to find a group of kids that are smart enough, uh, and then try to build it with them. Right. And, uh, the moment, I think it was Lee Marie from, uh, Kleiner mm-hmm.
Um, actually posted on LinkedIn saying like, Hey. There, there's this company, uh, this AI company looking for their first like business hire. Mm-hmm. Uh, I was like, okay, let's, let me talk to them please. Uh, and then when I met like Varun and, and Douglas and the team, I was like, this is definitely that smart team.
Uh, it's time to go in and build the, the company together. And this was like as podium was just starting to go live. Uh, pre-revenue, uh, trying to figure out where, where it was gonna make its first dollar. Um, and, and I was like, this is, this is gonna be a fun journey and I know this is gonna be big 'cause it's so early.
And I think, I think with any trend, as long as you're early enough and you have this, you know, smart enough group of people to work with, yeah. You're, you're probably gonna be successful. Um, or at least, uh, you know, hopefully the oddsly in that way. Yeah.
Selina: Yeah. What, what gave you that insight that this was, you were early on a trend, like how did you know this was going to be a big trend?
Jeff: Yeah, I think like with copilot. It was very obvious, uh, how much value it would provide. Mm-hmm. And a lot of times when you look at startups as investors, you actually want to just like, solve a problem. And the problem copilot was solving was like, uh, you know, the, the, the velocity of the code. Right. And, um, I think it was just so obvious that copilot was like one of the killer apps in the space.
Mm-hmm. That like when you join a company like Podium, you know, still trying to find. Where it's gonna do to compete and differentiate. But I think just knowing that there's a high value target there is, is the start. And then like, when you try to think about how you run the business after, that's where like you, you have to like, you know, put in the work, right?
But, uh, I think you gotta find the start and then go from there.
Selina: And for our audience who's less familiar with. Codeium Windsurf story. Could you share more about just what the company does?
Jeff: Yeah. Uh, so Windsurf is a, uh, a place where you build technology faster. Uh, it is where you actually write the code, and as you are kind of trying to come up with ideas to do more complex stuff, you can just ask Cascade.
In, in windsurf to, to build it for you. People used to count in terms of like tens of percents, like I'm like 10% more productive or 20% more productive. Um, and we used to write like half the code, uh, in, in, uh, organizations Now, like with these age agentic products, um, we're writing like more than 90% of the code.
In organizations now, which is absolutely crazy to think about. Um, I, I think like over the, the, the next, even like next year, next few months, um, probably a hundred times more code is gonna be written. And, and that is just like, you gotta think about what are the ramifications of that. Um, but anyway, I, I went a little too far there, but.
Palak: Yeah. Well, let's, let's go back. So you, when you started, uh, windsurf was called Codem, and back then it was focused on building a, I think coating extension. And then the company launched the windsurf IDE and then rebranded to be called Windsurf. From your vantage point, how do you think through that pivot?
Jeff: I think there was, um, you always kind of look at what the ceiling is before you do a pivot like this is, this was the second major pivot in the company. Uh, first it was a GPU virtualization to podium, right? Coding agents it's called of function, exit function. And then from the coding agents and extensions into like a an ID and you always look at like kind of what the ceiling is of each market.
And that's kinda what happened with the GP virtualization. There was like a ceiling of like how much revenue we thought we could get. And then when it, when we went to podium, we saw like the growth trajectory was kind of like linear. And you, you never want to be in a linear growth scenario. You wanna be in like an exponential curve.
And when you, when we were already experimenting with like, hey, what if we had these LLMs right to multiple files? What if we had them like actually execute? Certain things like the terminal, uh, like run, run a, a command to even, even either even like commit code or even like launch a browser or, or like launch a local host to see what you've built.
Like all these things were, we already had planned. Um, and then basically we were waiting for that moment to release windsurf. When it was all right, because it was like not a hundred percent there. Um, and then finally at the moment, you know, the LLMs got better. That's when we, that's when we launched Windsurf.
We were like right there prepared and we were the first, uh, IDE to do this, to actually write across multiple files, um, start from zero to one, very, very well even modify very large code bases. Um, because we have a lot, we have a lot of technology just to understand what is going on in the code base.
Palak: And around that time when you mentioned this GitHub copilot was the killer app a year ago.
Mm-hmm. You know, I think they went from like zero to a hundred million of a RR super quickly. What gave you the confidence to take on Microsoft? You know, it's like a David vis Goliath type of type of story.
Jeff: I think anytime there's an incumbent, um, one strategy is to either. Build a much better product or it's to just go where they aren't going for.
Right. So one thing we did was different deployment types. Uh, we had an on-prem offering. We had a hybrid offering, which other companies like they couldn't use copilot. Um, the other thing was a lot of companies were using GitHub and. There, there's like a, a copilot button when you open up GitHub. So it's kind of unfair that they have like this dis distribution advantage, but there's other companies that are not using GitHub.
So we were actually pulling what companies were like using Bitbucket and GitLab and just go after those companies and then what companies are like in the regulated space and go after those deployments. And that's how we initially got our first, um, like, uh, big, big companies, uh, in our, in our, um, customer base.
Um, and then over time, like we were always ahead of. At the feature base, uh, we, we came out with a chat feature way ahead of them. Uh, we came out with the, uh, obviously the ENT product way ahead of them. So just staying ahead of the curve, um, was another way to compete head to head. But it was still a coin flip when you go into these enterprise deals because they, they're just like, Hey, you guys are like.
10 people or 20 people, like how do we trust you guys over this, you know Microsoft, right? Yeah. So I think that was the tough part and that's why we focused on where like we knew they couldn't compete in. And then as we built that customer base, we started to compete directly head to head to head. Yeah.
Palak: Well that, that's been, that's been super interesting and you know, there's been a lot of conversations and, and thought pieces around the technical side of Code Co-generation, but here what you're describing is you took on this incumbent with really a unique go to market approach. Can you just talk more about that?
Jeff: I think one aspect was a lot of experimentation, so, uh, it was, at the time it was me and Ancho, just the two of us, and we, I think we were tackling 500 customers, so we had, oh my gosh, we had a new customer every 30 minutes that we would try to. Try the pitch again. We would ab test messages, um, and it was every 30 minutes for the entire day, for like 12 hours a day.
Um, and we would start forgetting like, have we met this person before? What was the, you know, history of this, uh, deal? And all we had was Google Docs and uh, Google Sheets. We were just like really going hard and we didn't let anything slip. Like there was no email slipped. Like one customer, five people. We had to win that deal no matter what.
Uh, we, because we were a startup with no revenue in this space, we had to get, we had to validate that we were building something that could get revenue. And I still remember the first day we had, our first customer was a thousand dollars, and we were so proud. We're like, I can't believe someone actually paid us for this product.
Um, and then of course, like that turned into like. 50,000 and then we had our, like first a hundred thousand dollars deal and eventually, like, uh, we had our first like seven figure deal. Um, and, and within like, I think three to four months from all this madness. Um, and, and it really just took a like sheer like intensity and, and, and velocity, like on the product side and on, just like not letting things go.
Selina: And how did your role in the organization change throughout all of this as you've built out a larger team?
Jeff: I think my job at the company was always putting myself out of a job. Uh, basically, basically I'd be like, you know, Jeff, you're like, you, you come in, you'd do marketing, you do finance. Okay, but you're, you gotta be doing sales now.
And it's like, okay, sure. And then it was like, okay, well now we have a sales team. You gotta be the first like, solutions engineer. And we call it deployed engineer. And it's like, okay, I'll be the first deployed engineer. And my, my calendar was again, completely full of just like doing demos and going through like kickoffs and everything.
And then it's like, okay, I think now that we've hired a deployed engineering team, you should like go try to run partnerships and then you should go do like the federal business you should go do. And basically like we had all these different lines of business, but they would just kind of throw me at kind of what needed to be done next.
And I would kind of like, try to make sure everything behind me was still properly maintained and everyone was, was still accountable. Um, so it got really crazy. Um, eventually, like all the. Entire sales stack, all the apps, all the systems, um, were all like, built by me. Like I think I'm still the admin on like all the, all the systems of the company.
Um, but basically like, just try, just charging forward Yeah. With like new ideas, new lines of businesses and, and making sure like whatever I leave behind is like, is still, still running appropriately.
Selina: Yeah. I guess when people say you wear multiple hats at a startup, you were, you were really, really living that experience.
Jeff: Exactly. Um, I think I've had everybody's job at, at some point at the company.
Selina: So it sounds like. You know, the business is doing well. Go to market is rocking, the product is starting to get traction. Why did the company start to think about m and a?
Jeff: It actually came as a surprise to me. Okay. I think we were.
Like we were fairly confident that the trajectory was looking good. Although I think one thing that we might have maybe overinvested in is like getting a lot of infrastructure, getting a lot of GPUs. Mm-hmm. Building a giant research team, because there were like kind of two things happening, which is like, we have this like very, very talented research team and we're building our own models and we're tuning models to like.
Just perform a lot better. Um, but it's very, very expensive. And then there's this other section of like, go to market and like this enterprise engineering team. And it's like also doing very well and it's like doing, it's also a rocket ship, right? But these things are kind of like. Kind of going into two different directions.
And I think, you know, I, I'm, I'm trying to, I'm gonna speak for the Varun, uh, now, which is like, hey, like we should, we should take that value that we have with this like, research team, because like having both at the same time is gonna be really hard to like, maintain the amount of revenue needed to, to support that.
Right. And I think I, it was a surprise when they told me like, Hey, um. You know, open AI is, is interested. Uh, I I was just like, whoa. Like this is, this is interesting. I mean, it's very cool that like probably the biggest company of our, of our time, uh, is, is interested in us. Um, and that, and that's how it kind of began.
Selina: The original home was supposed to be OpenAI. There are lots of rumors online about why, but from your perspective, what was the final nail in the coffin that broke that deal?
Jeff: I, I, I don't think I could speak too much about this, but what I can say is I. A lot of acquisition acquisitions now take a long time.
Um, it could take, you know, years, uh, especially in a situation where OpenAI is also has some relationship with Microsoft. And again, this is all reported in the news. Mm-hmm. I think there was definitely, um, multiple phases of, of crisises. Like for example, when philanthropic. Kind of got when something was up and they like closed access to the models.
That was like the first like major like, uh, kind of consequence of, of all the actions so far, right? Mm-hmm. And I, I can tell you the story there where it's like they said like, you, you only have a few days and, uh, we're just not gonna, we, we have to allocate this to other customers. And, um, basically I was on the phone like talking to a bunch of other providers.
And then it came down to like, we were about to shut off everyone from philanthropic just because we had no choice, there's no capacity, or if I could find it in time. And we were literally like within like three hours of shutting it off, um, where I found some from like one provider. And then, um, I believe like one of the providers was Bedrock and we were on the phone with their head, the head of Bedrock at mid midnight, and he was on vacation in Turkey or something.
Wow. And like we, we were like down to the wire, like we were about to shut it off like three hours later, even more. Um, so we, we, we got that, um, like capacity agreed upon. Uh, and then we, we were able to, not to shut off everybody, uh, in the morning, so it was like, it was like one crisis after another. But when you look at what I just talked about earlier, which is you have this like very expensive org you're running and there is like a lot of cash burn now.
Like what do you do if you cannot raise money anymore? So like, let's say like, Hey, I think this deal's gonna take like another year. Yeah. And it's like, wait, we're burning a lot of cash. Should we raise another round? No, we can't because there's this whole like, talk happening with OpenAI. Right. Um, so that, that's why is my assumption of, of what happened.
Um, but I, I can't say for sure 'cause that is like kind of behind the scenes.
Selina: Right. And so on July 11th, the news around Google's reverse acquihire plus the licensing deal that brought the founders of Windsurf and the core engineering team into Google. You talked a lot about the mixed reactions from the team, but I'm really curious, what was your reaction to all of this?
Jeff: I still remember, uh, they told me the Tuesday, uh, before, and I, I mean like. It was pretty shocking, I'll say. Yeah. Uh, I, I couldn't even, like, I, I was still pro, you know, it took me like probably two days to process actually Okay. Of what was going on. Um, and
Selina: did you go into the office during those two days?
Jeff: Yeah. I mean, I, I was, uh, trying to plan like a new like org structure mm-hmm. And how, how are we gonna like, come up with a messaging that is achievable and believable. Mm-hmm. I think that was probably the toughest part is probably a lot of people would lose belief after hearing the news. Yeah. And I think that that was tough.
And, and again, like it didn't really feel like. Like a promotion, it felt like kind of a, a punishment. 'cause it was like the weight of the world, uh, or the weight of like 250 people was on my shoulders. Yeah. Um, and, and it, it again, that was like a really, I would say there was, there was not enough time to be stressed, but I was definitely not getting any sleep.
Selina: In the flurry of all of this, you became the CEO. Did you volunteer to do this? Were you voluntold to do this?
Jeff: It was funny. Um, they told me like, Jeff, you're gonna be the CEO now. And I was like, okay, let me think of what I have to do. Right? Yeah. Um, and then later, I think the next day, a couple of the people that found out were like, wait, you agreed to this?
And I was like, what do you mean? I agreed to this? Like, I thought I had to do it. Like, I didn't think I had a choice. Right. Yeah. But at the same time, it also felt like I, I don't know who else would do it, um, just because there was not. People that knew the entire business, uh, and every individual as well as me.
Selina: Well, I think it's incredibly, um, courageous of you to step up and lead during this time when there's so much uncertainty in the business as the new CEO. What was your first order of action?
Jeff: I think the most important thing was increasing the probability of success. And to do that, it's increasing the number of options on the table.
Mm-hmm. And that could mean raising money, new, new money from investors. Mm-hmm. It could mean getting another acquisition, which seemed like. Kind of comical 'cause it, we just like kind of went through a, sort, a small one. Um, or it was actually like making sure everybody's on board with like, here's the direction of the company.
Let's focus on the things that we know we can win. But we're no longer going for exponential curve. We're, we're no longer VC-backed. We're, we're an entirely employee owned company, so we can do these things. But I don't think the employees like wanted that. Right. And, um, time was our enemy because recruiters started spamming, like multiple companies started spamming every employee at the company mm-hmm.
To go to leave. Um, we had customers who were starting to panic, right? There's customers saying like, I don't know if I'm comfortable with the future of this company now. Right? And this was all happening in the same day, in the, in that same Friday. Um, so time was our enemy and, um, having a lot of options on the table was the, the number one priority.
Selina: Okay. So it sounds like in this new role, you were weighing a couple of different options for the company. From your perspective, what was the North Star? What was the thing that you were trying to prioritize?
Jeff: So, it's funny, me, Graham Moreno and Kevin Lou, we were like the new board all of a sudden, and all of us looked at each other and we kind of just nodded saying like, acquisition is the best route here.
Right? And then we all just said, yes, okay, glad we're on the same page. Mm-hmm. And then we still had to go through all the, obviously all the options that we could. Um, so that was the North Star and. I was kind of spending, by the way, I was on the phone for probably nonstop after the all hands. And I was spending like half the time talking to employees, telling them, telling them, please don't leave.
Uh, we need you at the company. And maybe to calm down, like people were panicking and they were somewhere crying. And it was, it was a challenging time. And then the other half. Was trying to find like new investors or new potential acquisition targets. And like Graham Moreno was just like hitting all his context, getting as many intros as possible.
Um, and then I, I know like the, the, the legend is in, in, in the story now, which is like, saw the email from, from Scott Wu, uh, saying like, chat with the question mark. Um, and we also had this like text intro, I think from, uh, MongoDB. Um, intro to us as well. Yes. Um, but I, I didn't even see that text 'cause I was just on the phone like the entire time.
It was, it was absolutely mad. Absolute madness on that, on that Friday.
Selina: Yeah. And you sort of have two stakeholders at this point, right? Like you had your customers and then you had your employees. How did you balance the, uh, priorities of these two different groups in deciding to move into an acquisition?
Jeff: The funny thing is here, like without the employees, then it's all over. Especially like enterprise engineering team. Um, supporting the, all the existing products, making sure if anything had any downtime, someone could put it back up, right? Mm-hmm. That was like the number one thing in my head was making sure like the enterprise engineering team is there and they, and they believe that we can.
Move forward. Um, and it's not like I can say like, listen, we're gonna look for an acquisition and this is gonna happen. Right? There's just no way that you can say that and, and make it happen. Um, so that, that was like if you were to stack rank the priorities, it was like, make sure the company doesn't blow up and like that, you know, without the customers, you can still have the company, but without the company you can't have the customers anyway, right?
Um, so that, that's how I was like stack ranking the priorities on that day.
Selina: Yeah, I can tell that, you know, doing what's right by your employees is really top of mind for you. And one of the first things that you did was hold this all hands q and a for the team. There were a lot of mixed emotions at that point in time.
What was the hardest question you got?
Jeff: I think multiple people asked, okay, here's, there's probably two that, that I noticed the trend, which was one, it's hopeless, right? And like that, that was like kind of the one of like, everyone kept asking that question, and number two was, why not just give us whatever cash is remaining?
Right. And then people really didn't understand customer obligations and contracts and also liabilities. Um, it's not, it's not as simple as everybody thought, but I, you know, I can't say it in, in that manner. Um, so both of those questions were very difficult to navigate. And again, I was on the phone with, I was doing one-on-ones, and it was like, that's the same questions over and over, um, which there's no clear answer to, right?
But, uh, there, there, but then at the same time, I'm, I am trying to fix the problem. So it's like, how much, how, how much time should I dedicate to. To each of 'em. Right. And I was trying to split it in half, like half to employees and then half to try to solve the problem. Yeah. How'd you remain so calm and composed during all of this?
I just didn't have time to panic. I think that was all like, the only thing in my head was just like be relentlessly focused, relentlessly focused on just finding a solution. Uh, and, and there's nothing more important right now. Right. There's just, I for, I kept forgetting to eat also, uh, throughout the week.
So I lost like eight pounds that week, which is funny. Oh, wow. Um, but yeah, I just be completely focused on fixing the problem. Yeah. And I mean, it reminds me of the early days of like, here's all these customers and just focus on trying to close a deal. Mm-hmm. Um, just like as long as you have that focus, you increase the probability.
And again, the most important thing was increasing the probability of, of success here.
Selina: Ultimately, you chose to join forces with Scott Wu and the cognition team. You hammered this deal out in record time in about three days with so much going on. How did you navigate all of this on the fly? I
Jeff: think like after talking to Scott and Russell, um, they, they're, they're like, the whole situation was very compelling.
First of all, like the products did not have overlap. There's, we have a, a local synchronous agent, meaning you guide an agent to solve a problem and you try to get to that end result. They have Devon, which is a remote agent, and you could have like any number of remote agents, but you're kind of setting them off to do a very specific task.
And having the combination is very interesting because a like. We we're not really interfering with each other. Like we go into deals and we never really kind of compete with each other. Yeah. Um, but they, they covered very different services and they offer very different value propositions. And then on the other side, we didn't have any overlap on the team either, which was very interesting.
They had a core engineering team and they were trying to. Grow a very large go-to market team. And it just so happened that we had a very large go-to-market team and we had an enterprise engineering team, but we kind of had like lost our core engineering team. Mm-hmm. Um, so basically the teams really fit, the products, really fit, um, everything made a lot of sense there.
And we actually did have like a couple other. Conversations and a lot of threads going, but to me that was like a very obvious choice, um, the moment I talked to them. And from then on my focus was like, get that, keep that thread going, make sure that that, that is there is success there. Um, while there are still some other threads to like kind of close or at least hear people out.
Um, and there was like foundational model companies and there were, there were other AI companies that were interested, but it just didn't make as much sense as, as cognition.
Palak: And what were, what were the reactions of your customers, your employees at the time? Were you dealing with this all yourself? Were they in the loop?
And then when did you start letting them know and how did they react?
Jeff: Nobody knew until Monday. Uh, we, we had me, Graham and Kevin, um, working with, uh, Scott and Russell. We had like a dozen people in the office on Saturday all the way until very late night. Uh, and then on the next Sunday it was all the lawyers came in.
And I believe we had the lawyers from the Johnny Ive deal and from the scale AI deal. So these are like the best lawyers in the, in the bay. And we, we kinda just argued for like 24 hours pretty much. Um, but we got it done. I think like both me and Scott were like, you know, time is the most important element here and we are gonna get it done on on Monday.
And the crazy thing is, like we argued all the way until, I believe it was nine, 9:30 AM but we had scheduled the all hands at 10:00 AM Um, and we scheduled it earlier because we wanted everyone to stop panicking. The entire company was still calling me and being like, Hey, I have a family. I, you know, I got nothing.
Um, and, and I, I needed to find a job and basically I just, I had, I had to keep those calls going as this was happening. And, um, that, that all hands kind of like, made everyone a little bit like calmer. Um, and then again, we, we literally signed. The last piece at nine 30 and then we like got up and we just walked to the other building to, to do the all hands.
Yeah. That, that's how close it came. It, it was like just at the very last moment, uh, that things got done.
Selina: So from Friday to Monday, sounds like you got maybe five hours of sleep.
Jeff: It's about three hours of sleep. Three hours of sleep, yeah.
Selina: Yeah.
Jeff: But that one hour on like Sunday night was really helpful. It made, it made me, uh, good enough to talk on Monday, I guess.
So. Yeah.
Selina: Yeah. It's gotta be there for the 10:00 AM Yeah.
Jeff: Yeah. That was the most important thing.
Selina: So there's been many reverse acquihires that have recently announced. There's Meta and Scale, there's Google and Character, but the cognition m and A was PO Post Acquihire, which is super unique. And no other company has done this.
Do you think this is unique to Windsurf, or do you think there is a path for other companies to do this in the future?
Jeff: It, it will depend if they're, if the other company. If they can find a fit, right? Mm-hmm. Um, I, I think it also depends on like what their priority of the solutioning is, right? If they wanna run as a individual company, that is always an option.
I think with the Windsurf case, it was a little more, um, hectic, uh, compared to some other, some of the other companies where most of the team was still there, right? Um, I, I, I think like, again, it, it just depends if like there is like that perfect fit and, and if there's not, then it might be that the company needs to try to run by it as a standalone entity.
Selina: This type of reverse aqua hire deal where the founder and the rest of the employees have different paths. Should we expect this to become more common in the future?
Jeff: I mean, it depends on the incentive, right? If the prize is gonna be a $10 trillion industry and you have to have the right team to, to do it, yeah.
Um, then it'll probably continue. If the prize is that big, if the incentive is there, people will go for it.
Selina: And so if you are an employee at a company, what can you do to protect yourself in these types of situations?
Jeff: I think this is a very challenging question. Um, you have to like really trust. That the founders will at least leave the company in a state that it can do well.
Um, I, I don't think there is a way to like actually do get that knowledge, to be honest. Um, but I think that. This is, this could be an issue where the startup kind of game is changed, right? Because people are not sure whether the outcome, the, the probability of the outcome kind of goes down with the increased success of the company.
Mm-hmm. Which is actually a very strange to think about. Yeah. Right.
Selina: Yeah. It's like diverging paths between how the company does and how you as an employee working at this company, how you fare.
Jeff: Right.
Selina: So now you are about three weeks removed from the m and a saga. What advice would you give to leaders who might find themselves in a similar place?
Jeff: I think you just have to have that relentless focus on finding the right solution. And it, it does mean like not sleeping sometimes. Mm-hmm. Uh, unless of, of course, if you don't sleep and it makes things worse. Uh, but I think just if you, like, you have a lot of hours in the day. I know it's like, kind of weird to say, but like there's a lot of things you can do with those hours.
Whether it's like finding new leads or getting connecting with your network to find a solution. I think that is really up to that new leader to, to really just have that focus, to get that solution.
Selina: And in the future, what do you hope your team will remember you and your leadership for in this moment?
Jeff: I actually don't know.
I, I think they will just think we did the right thing, um, and we worked really hard to do the right thing. I, I think the leaders in general just should, should always focus on tr trying to do the right thing. And, and that means like not just for the company, but for the employees and. Yeah, it, it, it is a very tough scenario, but when you're in these tough situations, you just, you just need to focus on what is the, the best outcome for everybody.
Selina: Yeah. Is doing the right thing hard in these types of situations,
Jeff: it is very hard because the number of possibilities, um, the num, the, the right thing might be like a very small percentage of all those possibilities, and I think. There are ways to like rationalize, Hey, there's different paths where a subset of people will be okay.
Mm-hmm. Or there's a sub, there's a path where like everything's just like, like this is the best case that we could get. Right. There's, there's ways to think of that way. Um, I, I think it's very hard to find a solution where it's a win-win, win for everybody. I mean, I think in this case it, it happened to be that, but I don't know if that does exist for every time this happens.
Selina: Yeah. It's a complicated scenario where there's multiple different shades, but like finding, finding your path to what is the right outcome for the most amount of people. Like that's, that's what matters.
Palak: Yeah. Agreed. Well, I'm curious about the operating structure of the new business. Are you guys fully integrated?
Are you operating independently? What's it look like going forward?
Jeff: Yeah. Right now we're operating as two entities. Um, we are trying to find moments where integration makes sense. So there are like things in the marketing side where we're. Working alongside each other to, to push out some big campaigns.
And then on the product side, their team has come in to really kind of take over the latest updates. Um, I think in the first week alone, they shipped Wave 11, uh, which was already a pretty good release. And then we're already working on, uh, the next wave as well. Uh, so on the product side, pretty integrated right now.
And then the rest of the company is still operating a little bit as two, two companies. Um, but I think over time we'll try to figure out ways where we can make sure there's like more value, uh, together, working together.
Palak: Yeah, can you just talk a little bit more about Wave 11 and, and some of these future waves to come?
Jeff: Yeah, I mean a lot of it was like quality of life adjustments, like, and things like voice mode and, um, I think one thing with the waves to come is you'll see more integration with the remote agents. Uh, I think there's definitely a lot of technology that. That the Devon, uh, product has, such as understanding the code base and running long running tasks and getting that into windsurf.
So over time you'll see windsurf improve on the code base understanding, which already was doing well, but having these like things in the background, um, increase the understanding is gonna be, make a dramatic difference on the experience and then the ability to kind of set off these remote agents. Um, because a lot of times you, you don't wanna just sit there and, and wait for an agent for, for too long.
Um, but if you are, if you are able to do. To do that with Devon. I think that's gonna be a very powerful experience because there you, you'll pretty much cover, cover all the surface area. And, um, and with windsurf, by the way, um, we, we have, again, we are the most enterprise ready AI company in the space. Um, with all these deployment options, we, we even have FedRAMP, uh, and we have these, these, these capabilities that nobody, no AI company can actually serve the federal government right now.
Uh, that, so we're, we're just like focused on getting surface area, not only on the product side, but also on the market side as well.
Palak: So from what it sounds like the products have been. You know, very complimentary. How about the teams? Um, how complimentary have they been coming together?
Jeff: I think with any integration, there's always a little bit of friction of like, hey, who is, who's actually making the, the, the decision here?
Um, what's funny is that both companies did have a culture of like, you should always kind of find the ground truth. So I think it, it's kind of been a natural, again, a natural fit to, to actually have that friction. Just be, just because you can kind of arrive at. What, what should be the right conclusion, even with multiple decision makers.
Um, so again, there is, there has been obviously some friction, but it has arrived to some very good like outcomes as well.
Palak: Yeah, well, it's amazing that the products have come together so well, but the market feels just as competitive as ever. I think in the earnings, GitHub copilot announced they crossed 20 million developers open Air AKs, andro, launch Cloud Code Cursor.
They all have a lot of mind share. How do you position yourself in this environment?
Jeff: Yeah, I think this is not only the most competitive industry, um, in, in, in ai, and AI is probably the most competitive industry to begin with, but the velocity of growth is also kind of increasing as well. Like the number of users is increasing faster than we expected, that the amount of revenue is increasing faster than, than expected.
Um, but what we want to do is be, be that comprehensive platform for, for anybody. So any company should have no excuse that they can use Windsurf and Devon versus any other product. Um, and that, that. Comes to our ability to be thing in things like jet brains, uh, which for example, like cursor is not in. Um, and then I just talked about the FedRAMP instance where nobody's in there as well.
So our, our, our strategy is gonna be kind of trying to tackle a broad surface area where we can compete well in. And then in the meantime. Like I said in earlier, in the, in the talk just increased the product velocity where we are ahead of the, the competition. So having both kind of surface area and the product velocity, um, is, is a strategy to, to compete in this market.
And it, it is not doable with just any random team, but it is probably doable with, with the cognition team.
Palak: It makes a ton of sense. Product velocity needs to go up. I'm curious, as the model labs have launched products in this space, how do you think you'll continue to cooperate versus potentially compete with each other?
I think that's a really good
Jeff: question. Um, there're all, again, in the entire AI industry, everybody's starting to like grow their kind of like, there's starting an overlap into each other's, uh, kind of products. Um, one thing that we probably will try to do, well, there's actually two things. One is we wanna make sure the model optionality is a strength and not a weakness.
So if there is a model lab that actually only offers our model and tries to go big with that product. Inevitably there's gonna be a leapfrogging from another model, right? And by having the model optionality, that kind of positions us in a, where, a place where we'll always just have the best model, uh, that can be used for the, for the platform.
Um, on the other hand, like open source is catching up very quickly, and open source is driving down the cost to serve the models as well. I mean, just last month there, there's been like three very, very competitive open source models released to the market. Um, we can do a lot of different things, uh, with that as well.
Uh, and that again just comes down to optionality, but it, it also means like. These companies are not gonna be able to leverage any of the benefits that you can with these open source models.
Palak: Mm-hmm. Do you think it's important for you guys to train your own models in order to be at the cutting edge? In the research side?
I think we will find more
Jeff: efficient ways to do it. I think we will, there are ways to, uh, work with models to, that are not as expensive, uh, that are still very, very efficient, uh, in, in, in getting an edge against, uh, any of the model labs.
Palak: AI is really affecting the entire software development lifecycle.
You know, AI is writing code, it's reviewing prs, and you know, the whole DevOps stack is gonna change. How is software development gonna change in the near term and over the long term?
Jeff: I think in the near term is developers that are, that are working with these tools like alongside. Then as you will notice, like even in the next few months, it's gonna stretch across the other parts of the developer, like kind of job.
Like if it's a planning mode, um, like what are, what are we gonna build? Like what does this look like? Right? There are, there's a lot of like optimizations you can do there, like creating many different plans before you execute. Uh, which is something Devin is really good at. Um, on the kind of tail end of things.
Like, oh, like what about the testing side? How do you make this more, uh, secure? Or how do I just make sure like. We, we write a bunch of unit tests and make sure they're all being executed correctly, and then reiterate into the, into the platform, there are these things on both the front end and the back end that this, uh, these products will probably kind of start getting into.
And then. On the other, uh, hand, you'll have these people that are not developers, right, that are suddenly able to use these tools. Um, we, we are, we are already seeing this with, of course, the vibe coding tools, but there are like tech adjacent people that are able to just like, quickly prototype and build these apps and, and, and show this is what we, what we wanna build as a team.
And then, uh, actual engineering team will go in and, and make the modifications. But it's, it's getting to the point where there's only a few more features you need for. That prototype to actually be a fully functioning app. Um, so we might see that become more robust over time as well. Um, but then again, this is really where, where the market is gonna take us because like vibe coating and enterprise is still very risky right now.
Um, we did set up a team at Windsurf to do, to build a bunch of internal apps and it's been awesome. But when you are talking about like a company with hundreds of thousands of employees and there's like just a bunch of apps being built everywhere, it could get crazy, right? Um, so anyways, my point is you'll see.
All over the place, you'll see, uh, software development lifecycle, but you'll also see non-developers starting to contribute to what a developer does as well.
Palak: Yeah, and it's so incredible 'cause you've really been, you know, we talked about GPT two, like you've really been in this since the early days and then now you know, we're talking about all these apps being built within these organizations.
Is this kind of how you predicted it was gonna play out or what's kind of surprised you the most? The funny thing is the barometer of success
Jeff: at Windsurf was whether Jeff could actually contribute, uh, to the code base. And, and I, I actually think we, we kind of got there, um, but I think we're, we're now beyond that, right?
We're beyond to the point where actually these apps can probably build themselves and they can probably test themselves and they could even go bigger by themselves. And I think the models will keep improving Where. That, that'll just make it more and more robust. Obviously there are situations where like you can't just have windsurf build windsurf, right?
It, it won't work. Um, so there are like some challenges in, in more advanced kind of, uh, coating tasks, but, um, it, it, it is slowly getting there. And I do think there's like, um, the, the like an exponentially. Exponential difficulty curve too. So the, the, the closer you get to these like advanced things, things, it's like there's less data, there's less, um, like there's more uniqueness to building these things.
So it's harder and harder to get, um, more and more of the code written. So while we're at like 90% of code written now mm-hmm. That last 10% is, is, is very, very challenging to get to. Um, and our intent right now. Is to increase the predictability of understanding what the user is doing. Um, if they do like five things in a row, what is the sixth and seventh, eighth and ninth thing we think they're gonna do?
And the better we can get at predicting the user intent, the, the better we can get at building these advance tasks as well. So like, I think you, you'll see a lot of these like kind of advancements over, over the next couple months.
Palak: What do you sort of think the world or the job of engineering looks like?
Um, at the sort of end state once, you know, AI has kind of run its course,
Jeff: I mean, it's gonna change. Um, I think it's more about being an engineering manager versus a programmer. So, um, like organizing the things that need to be done, making sure they're well executed. Um, just managing the amount of code that is getting written.
Uh, again, like I said earlier, I think there's probably gonna be a hundred times, uh, uh, of code actually written. Uh, in general. So that would mean like organizations in theory will be way more productive. Um, and I think that, that, that transition's gonna take some time. I think people right now are just like, whoa, I just finished my engineering task and uh, I can just like, I'll, I'll just go watch TV or something, you know?
Um, but over time it's gonna be like the competition is building twice as fast now, right? And it's gonna be the point where like, okay, we can't just sit around because our competition's building twice as fast. We need to build twice as fast, but then these models get better. And of course everybody starts building faster.
So we're gonna see this like again, this, this whole talk, this whole talk track is about exponential curves, but like ex exponential curve of productivity too. Um, because you have to otherwise your com your competition is gonna do it.
Palak: Yeah. How do your customers go about measuring that developer productivity?
Jeff: We like, have a lot of different ways. Um, our personal way was like the amount of code we write, but it, that doesn't really capture like the velocity, but it does capture how much AI's contributing to, uh, to your code base through your committed code base. And I think over time it's gonna be about like how many things, like how many features are quickly being built, um, if you're using like Jira or linear, how many, how fast are you completing all those tasks or stories.
Um, and, and right now, even like for example, Devon, you can just like select a Jira ticket. Just say like, fix it with Devon. Um, and we're getting to the point where like, that might be so common, um, that you're kind of like removing all the things the engineer doesn't wanna do and you, you're giving them the things they wanna do.
Right. And eventually that could be also just directing an AI to build something that the engineer wants to build too. And it's already happening now, but it's just, it just takes time for like, people to get into that mindset to realize, oh, the competition's going faster, now I have to go faster.
Palak: Yeah. In, in that world, how do you think about that sort of.
Trade off between, you know, does a developer even need windsurf, you know, an IDE where they're editing the code directly versus, well, hey, I don't even have to do that. I can just go in linear and say, Hey Devin, you go do it. Or, uh, how do you think that's gonna play out? Yeah. I think over time, like IDs
Jeff: will go away.
It's gonna take a lot of time though, right? Yeah. And naturally more people will use these, uh, remote agents versus using the IDE. And that's why it's, that's why it's great that we are now with, uh, cognition. 'cause Devon and Windsurf can now work together and it, it's, it's probably gonna take a few years, but we're gonna see people like.
The, the amount of ID IDE usage probably like transition to more, more agent usage, right? Yeah. And at the end of the day, they're both agents. They're just like, one is in real time and you're guiding it. And then the other is kind of like you set off to do, to, to do the work by itself. Um, so you kind of need both today and probably for the next couple years, right?
Yeah. But that's what, that's what the trend is, is is all headed more towards agents.
Selina: So Scott has talked about how today we might be at 1.5 x productivity with coding assistance. In the future, he thinks we can get to 10 x productivity. From your perspective, where will we land on this, uh, spectrum by the end of this year, given how fast things are moving?
Jeff: I, I think some people are al are already 10 x. Okay. And it really is like a kind of like, do you have enough problems to solve?
Selina: Yeah.
Jeff: Right. And if you have enough problems to solve, how good are you at delegating these agents to solve them? Mm-hmm. And I think that's. What, what is going to matter? Like if there's again, a company that can help organize the things you can solve that agents can do, and you just click a button that it solves a hundred.
Mm-hmm. That means you're a hundred x faster, right? So if, if, if companies can reorganize, like their priorities and whatever they want to build or whatever they want to fix, um, and we're finding, Devon is great at fixing a lot of the legacy code stuff, right? Right. If you have a list of a thousand things and you hand it off like one person and they can do that in, in that year.
I mean, you're, that's a thousand x faster, right? Yeah. Like, so it basically depends on how you can organize things that an agent can do and, and execute on them and presenting that in into, like your, your engineering team. So are you able to contribute to the code base today? I, I am, but, uh, I don't know if they want me to. I'll, I'll have to ask for, uh, for permission.
Palak: Really amazing. Amazing. Every hat from every business role to CEO to engineer and maybe engineering manager one day. We'll see. Yeah.
Selina: So Jeff, we are investors and we like to dream the dream a bit. Mm-hmm. Walk us through how we go from today to cognition and windsurf becoming the a hundred billion dollar business.
Like what is the diabolical plan?
Jeff: I, I mean, you can just look at the amount of work, the amount of code that needs to be written. Mm-hmm. Um, there are like legacy banks that are still. Using like COBOL in the back, in the mainframe, um, there's a lot of, uh, code that needs to be upgraded like from Java, like from, I dunno, seven to 21.
Like, there's all this stuff that actually still needs to be done. And like I said, like when I say a hundred x of code is probably gonna be written, uh, over this next period of time, it's probably not even enough. There's just so much code that needs to be written. And then when you think about new products that are trying to get built, there's again, unlimited because like every company is trying to build new products at all the, at any given time, right?
So the di diabolical plan is just making sure that we are enabling all these companies to build at, again, 10 x or a hundred x faster, and that that is definitely worth the a hundred billion dollars prize.
Selina: Okay, so Jeff, we're now on the final question of the show, as you know, on the fly is about leaders making critical decisions that impact the future of their businesses.
What is a leader that you admire?
Jeff: I have, I'd say Scott Wu right now. Right.
Selina: Why Scott?
Jeff: I mean, this guy is, has built like an amazing team. He's gotten like, uh, I don't know if you guys know, like there's an IOI Gold medalist, uh, count. Mm-hmm. I think, I think he alone has three, but like the entire team has like somewhere between 10 and 20 gold medals.
Mm-hmm. Um, and we always talk about like, oh, like what does that mean? Is that impressive? Well, like, if you look at the cursor team, they only have bronze medals. Um, so he's built like an amazing team. And then, um, he's executed on an industry where it's like, hasn't really, he doesn't really have that much competition right now.
Um, a lot of these, um, agents, these remote agents mm-hmm. Um, are not like widely publicized yet, but the ones that Devin are doing are like, very, very valuable. And I think that this is where like he's found a, a spot where there's no competition. They're executing very high, very high velocity, and like in terms of revenue growth also executing very well.
Um, so again, maybe he might be the most like underrated, uh, CEO right now in, uh, in Silicon Valley.
Selina: Audience members, thank you so much for joining us on our very first on the Fly episode. If you would like to nominate someone for our future episodes, please comment or send us a message. You can reach us at our contact information in the descriptions below.
As always, all of you shared here are personal and nothing should be taken as investment advice.
