Limitless: An AI Podcast - The War for Windsurf: The $3B AI Coding Agent (OpenAI vs Google vs Cognition)

Episode Date: July 17, 2025

Windsurf has had a tumultuous few weeks. The AI Coding startup's $3 billion acquisition by OpenAI fell through, leaving its 250 employees in turmoil. We discuss the tension between corporate ambitions and employee welfare, spotlighting the ethical dilemmas in the tech industry. Following the drama, Cognition steps in to acquire Windsurf's assets, providing remaining employees with a second chance. Tune in for key insights on corporate responsibility and the harsh realities of startup life.------💫 LIMITLESS | SUBSCRIBE & FOLLOWhttps://limitless.bankless.com/https://x.com/LimitlessFT------TIMESTAMPS00:00 Setting The Stage03:55 How Does It Work?08:24 The $3 Billion Dollar Rug Pull11:48 What Led To This14:47 Why Are They Worth SO MUCH?18:50 New Acquisition!21:53 Who Really Wins Here?------RESOURCESJosh: https://x.com/Josh_KaleEjaaz:https://x.com/cryptopunk7213------Not financial or tax advice. See our investment disclosures here:https://www.bankless.com/disclosures⁠

So there's this company called Windsurf, which in the last 24 hours got stolen from Open AI and sold not once, but twice, for over $3 billion. And then everyone lost their minds completely when they realized that two of the co-founders were taking almost all of that money for themselves and leaving the rest of the company, 250 employees in the dust. And as you can imagine, Josh, this sparked a lot of outrage online, but it did end up having a few hundred and fairy tale ending when another company swooped in and bought the rest of the company and the 250 employees got paid. Josh, I'm exhausted. Can you tell me what's happening? Well, I'm actually not sure myself. This feels like an episode out of like a sitcom or a Silicon Valley episode where there's so much drama, so many twists and turns that seemingly don't make any sense. I mean, it was a quite I have a fun anecdote for you, Josh, actually. The office that the Winserve company was set up in,

guess. No, not Silicon. Silicon Valley. Silicon Valley. Oh, man, it's so real. Silicon Valley was such a testament at the time. I swear. They were ahead of the curve. But yeah, I'm actually, I'm excited to record this episode because I need you to explain

to me a lot of what's happening. I don't know too much about the recent news. I've been very much in the grok rabbit hole. I know a lot of the earlier topics. Maybe I could help set the stage before we get into it. So for people who aren't familiar, WinSurf was not always WinSurf. It was a company that was founded in 2021 called Exifunction. And it was founded by two MIT graduates initially focused on GPU optimization.

So they were a GPU optimizer company. They were doing absolutely nothing relative to what they're doing today. Three years later, they raised a Series C round at $150 million, which valued them unicorn status, $1.5 billion. And that was when they rebranded to WinServe. And WinServe was not a GPU optimization company. It was instead the company that we all know today, which is the AI coding companion, and cursor competitor, vibe coding, extraordinary.

Exactly, yeah. So that's when they kind of came into their final form, which was only last year. So the first three years of operation, this strange company trying to figure it out last year was huge. And then we get into 2025, where their annual revenue went from not a lot to $100 million plus, and I think this is where I'm going to hand it off to you

because things got crazy. They turned into a really big company. They started collecting valuable data. Their revenue went through the roof into the nine figures, And that's where this drama begins. Yeah. So Josh, do you remember around that time, it became a kind of trending narrative that AI coding or coders was going to replace software engineers. So companies were getting really nervous and software engineers were getting really nervous because they were kind of like consistently held in as like the highest regard of employee, right?

You had like knowledge workers, product managers and then you had like the wizards at the top that could like create software with their fingers. And then you had AI come along and be like, well, actually, we might be able to replace you in about six months. So all of these companies' valuations skyrocketed. WinSurf being one of them, right? Cursor was another popular one. And there was this massive frenzy for all the model creator companies like OpenAI, Clord, Meta, to try and acquire one of these companies so that they can have that feature set within their own product suite. And Open AI kind of came in guns ablaze and said, Winsurf, we have tried to buy cursive. it's not working, we need you, and we are willing to pay big bucks.

So they offered basically $3 billion, and it's written here in this Reuters article, but that's basically the headline takeaway, that they offered $3 billion to buy these guys out. And that was a huge price tag at the time, but it was what a lot of individuals thought, a lot of people thought that WinSurf and AI vibe coding companies was worth at the time. And the reason why it was worth this much was because it was like this magical tool where you could type words and turn words into fully fledged apps, games, tools, services, therapy apps, gym logs, whatever your mind could potentially think of. It's like turning chatGBT into like this like app maker.

And that, as you can imagine, is a billion dollar, multi-billion dollar industry. And you can kind of like see on my screen here, like there's this like demo on this website where you can effectively just type words in and say like, hey, I want you to create like a third. folder or a website that like kind of displays my profile picture and all these kinds of things. And it just, you can see it coding in real time on the left here, Josh. Yeah. And the interface is the thing that's different, right? Because when you're using chat GPT, it can code very well, but you only have a text box.

And it kind of spits out code, but then you have to put it into your compiler. And it creates this like very messy workflow. What windsurf did is they kind of took that whole workflow and compressed it down into one really powerful tool. And that's what we're seeing here, right? It's like, this is the developers come all in one, build your code, generate your code, debug your code all under one. Exactly.

So you can imagine that if you are a software engineer that learns how to use this tool or gets access to a tool like this, you can suddenly work at 10x your speed and output, right? So there's this term that's often used in tech, which is called like 100x engineer, and they're usually super rare. This tool turns everyone into 100x engineer,

and that means that you could hire probably fewer engineers to get your thing done, Or you could scale your company 100x from where it is right now just using existing employees and access to this tool, which is insane, right? So you can imagine something like this is worth a ton of money. But if you remember on a previous episode, Josh, we actually mentioned that OpenAI wasn't just,

didn't just get to where they were on their own, right? They had some pretty impressive backers behind them, and one of their backers was this little-known company called Microsoft. I remember them. You remember Microsoft, you know? They've been around a while, right? And some of the terms that Microsoft struck with OpenA.I. When they invested, I think, something to the tune of $16 billion, Josh, or is it like 24?

I feel like I'm off by like $10 billion, which is just like insane. Tens of billions. Okay, let's just go with casual tens of billions. Part of the terms that they struck with Open AI was that they get full access and IP control over OpenAI's models up until the year 2030. And also they get a significant chunk of revenue. that Open AI generates. And the final point,

which is what is about to cause the rift that I'm about to describe, was access to any and all IP that they may acquire as well, Josh. So as you can imagine, Open AI is buying this massively valuable company, Winsurf, for $3 billion, and their plan, presumably,

is to build an amazing product suite and earn tons of money from it. By Microsoft would be eligible to owning the IP of Winsuff. And Sam Altman, OpenAI didn't like this. So the rumor behind the scene was that Sam Altman was negotiating with Microsoft basically trying to get them off the cap table with that respect and not give them access to Windsor. There was a lot of back and forth, a lot of insider knowledge basically saying that Satya wasn't happy and that they couldn't come to an agreement. So the rumor was Satya walked away from the negotiating table.

Now, typically when an offer is made as big as this, $3 billion, it comes with some terms and conditions. and usually it comes with an expiry date. So if, God forbid, anything goes wrong, the deal suddenly becomes off the table. So, you know, typically anyone has kind of like an out. And what happened after Microsoft walked away from the negotiating table was that this $3 billion offer, Josh, expired on July 11th. So literally a few days ago.

And that's when all the drama started unwinding. Expiration, that is like a tremendous amount of money just to let expire. Right? I feel like if I was, I mean, put yourself in the shoes of a windsurf employee where you own equity, you are worth on paper now a tremendous amount of money. And you're sitting there every day waking up, looking at like the news like, am I going to get this money or is this going to respond? And then it expires. It's like an option. If you have a call option, it's just expired worthless.

You're left with nothing. So now the windsurf employee, I guess I'll play that role because I'm pretty clueless in this whole thing. I'm pretty sad. But I know my company's still worth a lot. So, like, what happens next to me as the windsurf employee after the AI deal falls through? I want you to envision this, Josh. Okay. So you're a windsurf employee and you've kind of been told that Open AI is going to buy you for $3 billion.

You go to bed one day. You wake up the next and you realize that the deal has expired, right? And then a couple of hours, yeah, it's expired. It's done. And then a couple of hours later, you open up your laptop and you see this tweet, Josh. Tell me about your reaction. Big welcome to Mohan Solo and others from the windsurf team joining Deep Mind. Now, if I remember correctly, the name of my CEO is the same handle that that was tagged in this post.

Exactly. Okay. You're good so far. So I believe I've just learned that the CEO of my company, as well as other, I guess, higher executives. You, the windsurf team. Me. I'm joining Google. I'm going to Google today.

You're joining Google, Josh. You're joining Google. Congrats. Okay, so you lost open A. Exactly. And now your question is, are you paying $3 billion? Well, what if I told you this, Josh? They're not paying you $3 billion.

They're paying you a little less. They're paying you $2.4 billion. Okay. How does that sound? So far so good. Yeah? I mean, at least you're better than nothing. You'll take that.

You'll take it, right? That'll make me multimillionaire for sure. That seems pretty sure. Sounds good. Except that there's one clause, which is that $2.4 billion, dollars is actually being used to buy your CEO that was just named in that tweet at one of his other co-founders and about four researchers on the entire company team. So this $2.4 billion is really just an aquire for the talent.

And they're still not buying my company. Yeah. So the long story short of this is Google announced that. they struck a deal with Winsurf, particularly the CEO, and agreed to pay $2.4 billion, but primarily to hire Winsurf CEO and one other co-founder and a group of key researchers, along with the non-exclusive rights, non-exclusive licensing rights to some of Winsurf's technology. So this is, as you said, you just got Zod. They just paid a couple hundred million dollars, took the talent out of my company,

and now I'm just left with the shell, which is devastating, right? Like I worked so hard on building this. I am the reason this value exists because I have been working hardcore engineering on this for years. And that seems really upsetting. Think about it. You are a founding engineer of this company. You're a founding employee of this company. Typically, you'd expect when you go from zero to $3 billion, you're going to see at least some of that.

Surely they're not going to concentrate over $1.25 billion for a single person. Sure, that's ludicrous, right? But that's unfortunately exactly what happened. So 2.4 billion for like five people and non-exclusive licensing rights for like some of Winsurfs tech. But the majority of the 250 people, its core business, the IP, the product, the brand, and the remaining team just kind of got left to the wayside. This was a pure play aqua hire. Yikes. And I guess we could kind of speculate on what led to this.

Like if you remember, we actually spoke about this briefly on the show. when Open AI made the offer to acquire WinSurf, immediately Anthropic and a few others started pulling out their contracts with WinServe. So they were making the product less valuable because they didn't want to give their data over to Open AI. And as a result, it lowered the valuation of the company. So when Open AI then steps away and they, WinSurf still has to deal with the lost customers of Anthropic and whoever else stepped away, well, suddenly the company is worth a little bit less. and now Google steps in, they're like, okay, well, maybe we get a discount because now your average revenue, your recurring revenue is way lower than it was just a few months ago. So there's

your 25% haircut. And then they're like, well, we don't actually want the business because the business is kind of failing. There's a lot of drama. There's a lot of, a lot of your larger clients aren't interested because they're direct competitors. So let's just take the talent. And that is such a a sad thing to watch your baby just get pummeled over and over again by these high executives. But it's on brand, right? Like we've seen this happen before Zuck is offering hundreds of millions to billions for employees. And now Google is doing the same exact thing. I think you just highlighted a really important trend, which has just been confirmed by these events, Josh,

which is the stars of the show for any AI company are its researchers, are its model creators. They're the ones that are selling for like multiples of a top NBA super sports star, right? Hundreds and hundreds of millions, potentially billions, for a single person. That is just an absurd amount of money and has probably never been seen the like from any kind of industry before, right? They're like professional sports athletes, yeah. Yeah, they're not worth their weight in gold. They're worth a hundred times their weight in gold, apparently, right? And then the second thing I think that's happening here, Josh, is Google got a bit antsy when they saw that Zuck aggressively poached Open AI's staff.

I think they hired like, what, 10 people. So that's probably somewhere around $2.5 billion spent if we want to believe the rumors of like hundreds of millions being spent for certain researchers. So Google's probably watched this happen over the last two weeks, which is an insane thing to say, by the way. So it happened over the last two weeks and thought, oh shit, we're next in line. We're the next best model provider for many a reason. They have access to all our AI papers so they can see who the best researchers that we have are.

And like Zuck's probably reaching out to all of these. So they saw this opportunity and jumped on it and decided to do the opposite and say, well, screw you, Zuck. I know what you're coming for, but I'm going to jump ahead of the curve and actually poach these guys before you do so that we keep bolstering confidence in our team

and our ability to build up the team. What do you think of that? Is that fun? Well, one of the biggest things I've been wondering is like, how are they actually worth $100 million? How are they worth a billion dollars for a single person? And I posted this publicly and I like did some digging. And the answer is really, it's still a fuzzy answer. But basically there are a series of difficult problems, most of which are unknown that are required to be solved in order to reach this goal. So in the case of training improvements. I mean, we spoke about Kimmy K2 yesterday. The reason that model performed so well is was because a novel breakthrough that a researcher discovered and then deployed into the codebase, and now they have a model that competes with Claude, but it's open source and free to use. So that's a

huge advantage, and that is a multi-billion dollar advantage at scale. So if you can pay one research to come up with a novel breakthrough like this and then deploy it to your code base, then suddenly you have gained billions of dollars of market cap, and it compounds because the scale is so large. And this exist all the way down the stack because when you're doing a training run, I mean, a lot of these companies have 100, 200,000 GPUs that are all focused on one single training run. And that means they can't be serving customers. They can't be serving queries to make the company money. It is a net spend that they're incurring to train this model. And in the case that an engineer didn't make the right decision or went down the wrong decision tree, that could like cost them many, many

billions of dollars, days, weeks of time. And in this world, those weeks cost a lot of money. So basically what I realized, what I came to consensus on is they're just giving themselves the best probability by having the most intelligent people on the specific topic of making the correct decision trees required to get to where they want to go. And that is why the stakes are so high. And that's why they're being paid billions of dollars, because someone like Ilya, who has been around for a decade plus, is probably most likely to make the right decisions, even though it's no guarantee. What you've just described is the Wild West,

and I love how you described it as kind of like an AI model, Josh. You were like, it's probabilistic. We don't really know which path it's going to take, but they're running it through this model, and they're estimating that the output would be like this, so they're spending hundreds of millions of dollars. I thought that was beautifully done. But Josh, if you remember at the start of this episode,

I said that it ended like a fairy tale. So right now, we're at devastation point. right? I'm deep in my nightmare right now. I'm living here. I've just seen my entire C-suite get poached for two and a half billion dollars. I got nothing. You just hustled for five to 10 years and you have nothing to show for it. You have a family of kids to feed and you promise them you're going to get them millions and nothing's happened, right? So here's what happens next. Do you remember way back when, and I think this is like a year ago, when the first AI engineer agent was released? It was this thing called Devin, D-E-V-I-N.

Do you remember that? I do. Yeah. Yeah. Do you remember the hype that was caused because of it? I think they released like this 30-second demo of this really slick AI agent. It's kind of like chat GPT, but it was just doing all the coding. And engineers lost their mind. They were like, oh, my God, I would.

There were two camps. One, I would love using a tool like this and it'll enhance me 10x or whatever. And then a vast wave of people online that hated it. and said, there's no way it's as good as it is. And then, you know, A16Z jumped in. I think they threw them like a $200 million check, classic A16Z. And there was a lot of hype around this.

That was like one of the clearest signals early on. That vibe coding was going to become a thing. Do you remember that? I do. So this company that created Devin is called cognition. And we saw this post come up literally the day after Google's acquisition. So imagine, you wake up one day, you realize you've lost the open AI deal. you're like, damn. And then a few hours later, you see the tweet from Logan Kilpatrick from

Google saying, hey, we've bought you for $2.4 billion. You're related until you realize that you're not getting paid any of that stuff. So you go into like peak depression. And then the next day, you see an announcement from Cognition. And this is the official announcement on Cognition's website. Cognition acquires windsurf. And it's this joint announcement from the CEO of Cognition, which is Scott Wu and one of the co-founders and by the way, this is the co-founder that didn't get included in the $2.4 billion

dollar win-served deal. Yeah, yeah. Well, calling him a co-founder is kind of like a bit of a kind of like graceful title to give him. But his name is Jeff Wang and he was the recently appointed CEO after the two co-founders just ditched him, basically.

And they were like, okay, Jeff, you're head of BD, but we need you to step up to CEO because we're going to bounce and take this billion dollar check, but you're the new CEO. So this guy, Jeff Wang and this CEO, Scott Wu of Cognition, struck a deal to basically take over the remaining IP, product, brand, and remaining team, most importantly, 250 individuals,

all have families, mouths to feed, whatever lives to live, and include them in this package deal. So this was the major announcement that kind of like swept the social media realms and made everyone super happy and elated by this and, you know, gave them their happy ending. What was the total cost that they paid? Did they announce that? That's a great question. Let me pull up this tweet of this guy called, oh, Harry is his title is Harry digresses. It wasn't official how much they got paid for, but there was some really amazing deal terms in this, Josh. So I'll highlight a few things here. The first one is this. So not only were all 250 employees that were

behind going to get paid, but they were going to have all their cliffs removed and have accelerated vesting on all that equity. Amazing. Okay, so they're making things right, or at least as best they can. It's not the $3 billion offer initially, but they're doing the best they can. And I mean, as an employee of WinSurf, that feels really good. It feels really good. But then you're wondering, you're wondering, well, how much am I getting paid? Like, how much are we getting acquired for? Yeah, let's say I had 1%. I mean, that's $30 million I was about to bag. And now I'm at So where are we at now? It might be a little less.

But he basically does this rundown and estimates it to be around $300 to $400 to $400 million, which is a significant drop from what you are going to get paid. But it's also not nothing. The other way I was thinking about it is it's probably all in cognition stock, Josh. There's no way they had that much cash in hand. Yeah, just cash sitting on hand. Okay, so my $30 million is now $3 million in. stock for cognition, which is a private company that is an IPO, not even liquid.

Oh, yeah. Lots of tax implications. Okay. I mean, at least I have a job, I guess, is where I'll end up on this. It's like, cool. I mean, now that I've recounted to you, Josh, you're still screwed. At this point, I'm just happy to be employed. I'm just glad that I still have some sort of a job.

I'm hopefully collecting some sort of paycheck. One day, it is hopeful that the stock will. be worth money and then I'll be able to sell it and cash out. But man, yeah, what a roller coaster. And it kind of sets this new paradigm, right? Where like you can cut just, you could kind of rip the value out of a company and then just kind of throw the rest to the side. It's just like really interesting ideological debate you could have as a founder of a company whose fiduciary duty is to like serve right by the shareholders. And in this case, your shareholders are the employees that you've given stock options to. And in the case that they're just going away.

And they've destroyed the value of the company. They've extracted the value of the acquisition. They have kind of just left the shell of a company and left the people who are responsible for putting them in that place on the side of the road. And like, that doesn't feel very good. But there's really nothing legally preventing them from stopping that. It's really just an ideological debate. In this case, it doesn't, I mean, I'd be pretty upset.

Sure, from one of the founders, I just made a billion. I'm now a billionaire. Like, that's cool. but for everyone else, it kind of leaves a bed for the majority of people. Yeah. Yeah, you're screwed. I don't know whether it's because of all the time I've spent in Web 3, Josh, but I've kind of grown numb to all this terrible, appalling behavior,

unequitable behavior. And so I kind of look at a story like this and I'm like, oh, yeah, he got paid billions and left the entire company and team to dry. Yeah, it happens all the time, I guess. But, like, I'm just, I'm noticing. this trend, particularly within AI, that there's this more Wild West kind of degenerate behavior when it comes to acquisitions. Like, people are ruthless. And I don't know whether this is, maybe it's my naivety when it comes to like VC deals and acquisitions and mergers. Maybe I just,

I just don't know. Maybe this happens all the time. But this seems like the most extremely skewed version of this. And we often refer to it on the show as like the Game of Thrones. And I don't think there's a much better description than that, where it really is cut throw. And this happens a lot in traditional venture and acquisitions, but the scale, I think, has never been greater. And that's probably what stands out among all of these headlines is just how much money is being made, lost, and thrown around to position yourself better to get to this point of AGI. Could you have a tweet?

You had a post here. It looks funny. Just to round us out, Josh, just to, like, put a bit of cherry on this chaos cake. Well, the TLDR is like, everyone hates Varun Mohan, the former CEO. who took the $1.5 billion check or however much it was. And everyone loves Scott Wu, the CEO of Cognition Lambs, who stepped up,

gave the 250 employees a salary and a lifeline and accelerated vesting. So no matter the size of the money, shout out to Scott Wu, he didn't need to do that. And then, you know, Varroon, I wonder, you know, whether his reputation survives this. I can see the reasoning why you would take the big check,

but also, like, leaving all your employees out like that to drive without even giving them, I don't know, a couple mill is kind of insane. But here we off. So what's the price on your reputation? Like, would you take a couple billion dollars to just go work on AGI and let the me do a few for a couple of people that would do it. I know a lot of people. I would need to be in that position to that.

I'd probably do that, right? Like a couple billion dollars, I don't know. But also, yeah, it's tough. Well, it's funny because at this graphic, it's a win-win for both of them. Like, Varun is, or I'm not sure how you pronounce his name, but he's hated. The CEO of WinSurf is not liked, but he's rich. And Scott Wu is loved. And now he has a bunch of employees that really want to work hard to make this work for him

and really want to add value to his company and can go to sleep with a good conscience knowing he's done well and he's done right by these people. Whereas our WinSurf buddy cannot do that. So maybe he will just be, one of them will be crying in their $100 million mansion while the other will be surrounded by people who love him dearly. Yeah, I think that's pretty much it. An absolutely outrageous, but weirdly normal day in the world of AI, Josh. We thank you guys for listening. We are trying out this new format. We've said this before where we kind of talk about trending topics and be or try

and be the first guys to talk about it and put it in front of your feeds. We are welcoming any and all feedback. So, you know, any kind of comments or DMs, it's all open. Go for it. and if you could like, subscribe and share it with all your friends or anyone who you think would be interested in this kind of thing, we would be eternally grateful and we'll see you on the next episode. See you on the next one. Peace, guys. Peace.