> Windsurf began in 2021 as Exafunction, founded by MIT graduates Varun Mohan and Douglas Chen. The company initially focused on GPU optimization before pivoting to AI-assisted coding tools, launching Codeium, which later evolved into Windsurf.
> Series B (January 2024): $65 million at a $500 million valuation.
> Series C (September 2024): $150 million, led by General Catalyst, at a $1.3 billion valuation.
> May 2025: $3 billion acquisition from OpenAI
I wonder how much of the value is really from the model or the tooling around it. They all use the same models (mostly Claude, others have been horrible and buggy in my experience). Even co-pilot agent mode now uses Claude. The editor has their own LLM (?) that does the apply since LLMs often return snippets. They work well enough on Cursor. And then you have the auto-complete, which I think is their own model as well.
But the main value from me is from the agent mode and 95% of the value is the underlying model. The other stuff could be more or less a VS Code plugin. The other benefit is the fixed pricing. I have no idea how much 500 calls cost if I were to use the API, but I expect they're probably losing money.
2) OpenAI recently raised 40B from SoftBank and others.
3) Windsurf is getting roughly 1% of OpenAI's valuation.
OpenAI needs to keep moving fast to outpace MS, Google, and others -- and I think we can all agree that agentic coding is a major trend -- that is likely to keep growing really fast -- and super high leverage in that the folks doing the coding are well paid -- and more likely to be early adopters than any other field. (e.g. if openAI wants a fast way to grow beyond $20-$200/month, owning a tool like windsurf is a good move)
Some folks have been speculating the cash/equity split. I'd be confident whatever number they arrived at de-risks things for windsurf, and preserves the right amount of cash on hand for openAI.
Even if OpenAI is burning 10-20B a year, with the recent raise would buy them between 1-2 years, and given the pace of AI development that's a pretty long time.
I’ve been a WindSurf customer since day one. It was my first true AI agentic experience.
[Dev mode]
While working on Alembic migrations I broke one of my migration files. After an hour of manual debugging I handed the task to WindSurf. It methodically checked every config file, applied the migrations one by one, and narrowed the issue to a single file. It rewrote the migration, verified the fix, wrote tests, ensured everything passed, and opened a PR. I reviewed it and it worked flawlessly.
Regarding the acquisition
I don’t understand why OAI would pay $3 B. The team is strong, they have lots of data, and the agentic flow is great, but all of that feels commoditized.
Claude Code launched two months ago and I prefer it to WindSurf, Cursor, and Aider. Augment Code also ranks above WindSurf for me.
If I were in Sam’s place I would have doubled or tripled down on Codex CLI.
Just my 2 cents.
I was involved in an M&A once; my role was to evaluate the technology and determine how long it would take us to build a competitive product. If it was less than some X then we’d build it, greater than X and we’d buy. The function for X was not clear to me from my perspective; it had legal fees involved, etc.
The person leading M&A said an intangible aspect of the price is what it does to the adjacent market. If the leading product A is valued during a raise at $Y, and you buy the next best product B at 1/10 that, you cause future issues with raises for A.
That's a really interesting thought, I'd love to get involved in software PE/M&A on the technical analysis side but I don't have the academic pedigree for it (it seems every shop that does this work is 90% Ivy and Ivy-adjacent universities and FAANG-level work history).
So if I'm understanding your point then part of the value in paying $3B for Windsurf is that it acts as a pricing anchor on future raises (and presumably acquisitions as well) for competing products? So Cursor is less likely to raise at a $30B valuation if Windsurf is 95% as good and just sold for 1/10 that.
I would also think that a critical component of X there would be the opportunity cost of time spent on building in-house while competition chugs along.
> WindSurf, Cursor, and Aider. Augment Code also ranks above WindSurf for me.
Bring on (a lot) more competition! I am waiting for the point where "Simple Pricing" (Augment Code has that on the pricing page) means fixed pricing; Simple is NOT '600 messages included' because it's impossible to know what the ROI from those 600 messages is, so it's very far from 'Simple' (many of those prompts will deliver nothing or, worse, having to rollback because the agent produced garbage). I know it's not sustainable, but the only thing that will keep me not jumping from one to the other, signing up with different emails, trials, coupons etc is if they will lose the restrictions on usage. They will, as they have to compete and it's worth it seeing this acquisition; losing 10s of millions a month to get/keep people and getting nice growth is what they do to get the billions. So bring it on!
If he's (Sam) trading equity on a grossly inflated OpenAI for the acquisition then he's likely not paying real money for the company and thus he is expanding his roadmap for cheap.
I don't think that's the case here. Windsurf wasn't leading the agentic coding market. They were doing a decent job but others are bigger. Cursor has the brand recognition and Claude is getting a lot of recognition too. MS has github copilot which is still a good brand and Google has been catching up with Gemini.
OpenAI has a new thing called codex but it isn't very good yet. I tried it and it's super flaky. Lot's of errors and it gets stuck when that happens. OpenAI needs something good urgently because agentic coding is the key AI feature right now and the blue print for non coding agentic solutions later. Cursor is probably too expensive currently and windsurf looks like their models are a bit better.
So, OpenAI gains something they don't have: a credible developer option with an active user base and some core IP in the form of training data and know how as well as custom models that they can fold into openai.
3 billion is a lot but not if you consider that world + dog in the enterprise world will be spending big time on AI subscriptions for their developers. This stops being optional in 2025. Millions of developers will be on paid subscriptions permanently very soon. If you start a new job you can expect to get a laptop and a paid subscription to whatever is the agentic coding tool of choice in your new company.
OpenAI wants double digit percentages of that revenue. 1M users paying something like 50$/month would amount to 600M revenue per year. I think the prices will go up and the amount of active users as well. Reason: as these tools are getting better they start saving non trivial amounts of engineering time. At that point you have to value the tool in terms of developer cost. Not 1 to 1. But it's worth a sizable portion of that.
I work in a small startup as the CTO. This is an no-brainer for us. We're cash strapped so we only spend on important things. This would be one of those things. We're doing things I previously would have needed to expand the team for because I would have had no capacity to do those things in the current team. So, in terms of value for money spending on these tools is easy to justify.
I get lots of people are skeptical about AI stuff here. But I would say that a lot of those people suffer from a short term focus and bias. Three years ago none of this stuff existed. Now it's a multi billion$ market that is set to grow rapidly. Stuff is getting better at a very rapid pace. Just stating facts here. 3 billion is a bargain if Openai can make this acquisition work for them. They are buying time to market here. They don't have a year to figure it out. In a year or so this market will be carved up and locked into hard to change year long SAAS contracts. At that point getting people to switch tools will get harder and harder.
> OpenAI has a new thing called codex but it isn't very good yet. I tried it and it's super flaky. Lot's of errors and it gets stuck when that happens.
I agree with this, not sure the experience of everyone else but I felt like Claude Code is more useful.
Meanwhile, I'm keeping tabs on Aider and open-codex, what other options are there?
Thanks for mentioning open-codex. Did not notice that there is a codex fork which is open to other models (update: totally missed that original codex allows that too now). How do you like it? Especially in comparison to Claude Code?
Not replace but it allows me to scale what we do for things we previously would have dropped because it would require growing the team, which we can't really afford. It's a case of getting a bit more out of developers in terms of quantity and scope (mostly this) of what they do. Not a full developer but enough for it to be meaningful. But it's not nothing either. Worth paying for. AI is a lot cheaper than a developer is so I don't need to replace my developers. I prefer people that are multi disciplinary and able to pick up new skills as they are needed. Agentic AIs are good for that because they give you enough to work with that you can get productive with whatever you need to wrap your head around in little to no time.
Companies can be a bit slow to update their hiring processes to their needs. But good developers should be ahead of the curve in any case. For this, just be proficient with the tools.
Be ready for the inevitable interview question "so, AI ... explain me how you are using it and what you are doing with it?". Much easier to answer that question if you have some meaningful time of routinely using this stuff behind you and can articulate what works and doesn't work for you.
And if they don't ask, that's actually a great question to ask back if you get the opportunity "I've been using agentic tooling, how are you guys using that a <company name>? Also I would like a subscription to <my favorite AI tool> if I work for you". Stuff like that makes you stand out as ambitious and interested in the future. There are of course going to be places that maybe don't like that. But then ask yourself whether you'd want to work there. So, either way, you learn something.
I would look forward to the next 20 years and not backward to the last 20.
The whole frontend/backend distinction did not really exist until the web. And infrastructure is definitely something that should be automated far more than it currently is. If it needs babysitting by a team of devops, you just created a lot of work rather than automating/solving it. Tedious and repetitive. It has "AI will make this a lot easier" written all over it.
So, just be ready for the ambition level to be raised for developers. Learn to build the whole system, not just bits and pieces of the system. Lean on AI to get stuff done and figure things out. It's all just code. None of it is really that hard. But it can be a lot of work if you do all of it manually.
And let's be honest, agentic tools are showing promise and great progress but they are nowhere close to independently working on existing code bases. That's not how I use them. But they are great for problem solving, debugging, prototyping, exploring some new languages and APIs, and generally taking care of more tedious coding tasks.
Because they have users and OpenAI has seen the massive drop off in coding usage since Claude Code came out. My personal Chatgpt decline is at least 99%. It’s also 1% of their current market value. So not really a big deal.
They're paying $3 billion because money is hyper plentiful for OpenAI at present. Basically because they can. Money isn't their problem right now, it's not a scarce resource (maybe it will be in the future of course). They're trying to capture and lock-in, so as the hurdles and regulations go up they're one of the huge winners left standing.
Try replacing Uber today, it's impossible. Nobody is going to give you billions of dollars to try to do it. It'd be an absolute nightmare to attempt it.
Uber has already been replaced, at least in some parts of the world. We recently went on holiday to Malta and on check-in the hotel staff told us not to bother with Uber, Bolt worked way better and had more drivers (Bolt is a European Uber competitor based in Estonia).
So we signed up for Bolt and sure enough drivers were plentiful, the app worked great and there was no downside over Uber. I'll certainly be trying it again in future in other markets.
The reason Uber invested in self-driving cars for years is that otherwise they have no sustainable edge. It's just a taxi company, which is a low margin business. People who can make slick mobile apps are plentiful and it takes a minute or less to sign up for a new service. Uber grew to its current size by buying market share using investors dollars, which was always a time-limited strategy. Once they started having to turn a profit prices rose and their edge over their competitors was lost.
Uber feels like such an apt comparison to OpenAI to me. The service they provide is obviously going to be absolutely huge, but no guarantees at all that they’ll win it or be last man standing. I don’t see a world in which generative AI doesn’t continue to be a massive disrupting force, but no particular reason to think Anthropic or OpenAI will still be independent entities in a few years.
I’m even more bearish on Uber than I used to be, as someone who’s used Grab and Careem and Bolt extensively, and seen Uber have to beat a retreat from SE Asia. If their more nimble competition get a foothold in the US they’re toast.
I feel like the value-proposition of Uber was three-fold.
1. Solving a pain-point of many people re: hailing a cab, via an app that works everywhere.
2. Using VC funds to (initially) pay drivers more than you, the customer, were paying them.
3. Ignoring local regulations and passing the savings/convenience on to you.
1 is nice but I don't think they established much of a moat (both drivers and customers are willing to use multiple apps); 2 isn't sustainable in the long-term, and they failed to leverage 3 to establishing a permanent right to operate as they had been in most markets.
I think this makes Uber an even more interesting benchmark for other unicorns, since besides "solving a real problem without establishing a moat" they are also often burning through VC cash to prop up their business model while ignoring some laws which they may not be able to get away with ignoring long-term.
1 especially is a social function. Having to have a million different apps is terrible, but if there is too much competition for drivers it's inevitable to churn through apps because of marketplace pricing and rent seeking on all sides
talented and smart folks for sure but can't not notice how much luck it is especially because its like 100% just better models. Windsurf raised a ton of money and then said they pivoted which they had millions raised to just do something completely different that likely wouldn't have been easier to raise for. Even in an interview with the cursor founder he kind of dumbly rambles that they launched and then basically lost a ton of traction until GPT4 came out. They have some core features like autocomplete but I'm struggling to see vision other than getting training data for iterative dev is a partial moat compared to just seeing commits and final code bases.
Training data is almost certainly their main reason for this acquisition. Users themselves and the models they use don't really matter. What matters is their interactions with the models. Especially if you're trying to build coding agents that will be marketed to companies for $10k a month. OpenAI is going for the industry B2B opportunity here, not consumers or end users.
5 years ago if you said coding tools would be worth in the billions in value it would of been surprising to most people. Dev Tools were the thing you could never get a company to buy for you or were just free for most people. Interesting times.
Dev tools are still very hard to sell (I know, I have a dev tools company). Claude Code, Aider and Codex are given away for free. What people are buying is access to proprietary general purpose models.
Yeah, in the recent Lightcone Podcast episode, Varun was talking about how they have a lean eng team but large sales org. I thought that was super interesting for a dev tool since I was expecting a dev tool to involve bottom-up sales to the dev instead of top-down sales to a leader like a CTO or VP of Eng
That's a bit like saying having access to Google is as good as being Google.
All they really see as a model provider is little fragments of the picture, like trying to reconstruct the Mona Lisa by knowing which paint swatches Leonardo used.
In other words, they only saw whatever Windsurf sent as context with a "fix the bugs" prompt stapled to it.
By owning Windsurf, they see the entire source code of what's being built, all the time, plus how the model is interacting with it.
There's a massive amount of value in what happens client-side, and behind the scenes. The "director's cut" of context.
So just put up together a comparable VS Code based AI IDE in a couple of months and bundle it together with the ChatGPT subscription? They'd get loads of users very fast..
I think it is exactly this. There is no doubt whatsoever OpenAI can do this, but they decided not to. The reason, I think, is that they don't want to be a couple of months late. In other words, they spent $3B to save a couple of months.
There's much more to be gained if you also have the client side of those interaction. You can get signals from "accepted" completions/plans/etc, number of edits made to those completions, how users use context, what was passed in context from a code base, and so on.
And that's just on their models. They'd also get (at the very least) signals on their direct competition, if not straight up prompts+completions as well.
The right time and the right place, plus they did the work, ofc; but I'm sure 80% of this site has worked as hard as, or even more, than what it takes to clone VSCode.
I'm jelly. Very rarely you see in history someone lucky enough to be riding the absolute top of the wave. Even OpenAI took about decade to cook their breakthrough product.
That was my thoughts too. No text editor is worth $3B, and probably not even VSCode is. So I think this deal was about buying more customers/users and buying "relevance". OpenAI lost it's monopoly and they're worried they might become irrelevant so they basically just purchased something popular to remain relevant.
I'm skeptical about this VSCode fork commanding a $3 billion valuation when it depends on API services it doesn't own. What's their moat here?
For comparison, JetBrains generates over $400 million in annual revenue and is valued around $7 billion. They've built proprietary technology and deep expertise in that market over decades.
If AI (terminology aside) replaces many professional software engineers and programmers like some of its fierce advocates say it would, wouldn't their potential customer base shrink?
Professionals typically drive enterprise revenue, while hobbyists—who might become the primary users—generally don't support the same business model or spending levels.
Part of what you're missing is that OpenAI needs to justify its own overinflated valuation. They raise money on the premise that an AI-native company can and will outcompete giant established players, so lowballing Windsurf would run counter to the narrative they're selling to their own investors.
The article also doesn't say that it's $3B in cash that OpenAI is spending. They might be giving Windsurf $3B worth of OpenAI shares - paying an inflated value for Windsurf with their own inflated value.
OpenAI just had a fundraising round that put them at $300B. Maybe they're just giving Windsurf 1% of OpenAI. Maybe they're even giving less than 1% - if OpenAI was worth $300B at the end of March and $150B last October, maybe they're worth $400B now. Maybe Windsurf is getting 0.75% of OpenAI that's "valued" at $3B.
> OpenAI just had a fundraising round that put them at $300B. Maybe they're just giving Windsurf 1% of OpenAI
That is the most hilarious maths I have ever seen, if this is true then it's maybe the biggest "holy fuck it's a bubble all the way down" I have ever seen
They built all of this assuming VSCode was a solid foundation for the next 30 years and I've completely undermined VSCode's technical foundations. Their castle is gonna sink into the swamp...
$400M in real revenue versus $300M in annual recurring revenue (ARR) are totally different things.
Real revenue is money actually earned, while ARR just multiplies one month's subscription revenue ($25M) by 12, ignoring customer churn.
Startups love flashing ARR figures because "$300M ARR" sounds impressive, but without knowing churn rates, they might never actually collect that full amount.
Unless you have reason to believe the revenue is declining in recent months or will decline in near future, ARR is a better metric. last year real revenue made sense only for low growth companies.
Cursor just lost access to the extension marketplace and key proprietary plugins that they were using against Microsoft's terms, Windsurf has been eating a chunk of their mindshare, and Copilot is catching up.
That's three good reasons to believe that lots of people will be cancelling in the next months unless something changes.
The entire reason OpenAI has a high valuation is the expectation that AI will get a lot better in the next few years. If that happens, building a clone of Cursor/windsurf should be trivial. The only reason you would buy windsurf today is to either pump up the bubble OR use it to increase your market share of developers by taking users away from claude
I feel jetbrains is squandering an opportunity here. Cursor is significantly easier to build then any IDE in the jetbrains ecosystem. The technology jetbrains is very hard to replicate. While the technology cursor uses should be trivial to replicate.
If jetbrains can combine there IDE technology with cursor technology, that would be ideal.
I think the problem is jetbrains tech is sort of already very biased in a certain direction and it's hard for them to pivot as fast into this new AI direction.
JetBrains launched their cursor competitor a few weeks ago.
I prefer Claude Code still because it has access to more tools - Junie seems unable to fetch URLs and do other things. But that's a tiny gap that JetBrains can close quickly, and the Junie UI is quite pretty. Plus, inside the IDE they can equip the model with far more advanced tools than Claude Code will have from the CLI: inside Code Claude has to explore the codebase by banging stones together with ripgrep, whereas in the IDE it can be equipped with tools to access the indexes and navigate around like a human would.
In theory, JetBrains should be able to compete very strongly in this market. Their single line completion model is already excellent.
> If jetbrains can combine there IDE technology with cursor technology, that would be ideal.
Just give them some time, they're not stupid. I'd drop Cursor in an instant once JetBrains catches up, because IntelliJ IDEs are just a way more powerful.
They're giving out 1 month free if you're paying for their IDEs already. I've tried it last year and it was very limited, not "agentic". Now they've launched an agentic version called Junie and also gave another 1 month free, and I've tried it again.
It's a nice improvement over the last edition, but still quite not "smart" as Cursor or Windsurf. The agent seems too shortsighted compared to competitors: it may stop looking for files or making edits sooner and you're left with code made with incomplete context (that does not work or just doesn't address your needs). It also does not fix linter/compiler errors from its own output code before finishing, unlike Cursor.
"I'm skeptical about this VSCode fork commanding a $3 billion valuation".. Nothing to be skeptical about. The market has spoken. It was worth 3b to OpenAI. Companies arent worth a vague notion of what "value" someone in an armchair thinks they might be worth, they are worth what people are willing to pay, and OpenAI paid.
> If AI (terminology aside) replaces many professional software engineers and programmers like some of its fierce advocates say it would, wouldn't their potential customer base shrink?
This is such a good point. The best reply available to the AI hype-men would probably be that LLMs "democratize" coding and therefore that even more people will use IDEs in the future, but that sounds like BS to me -- not unlike AI/hype itself.
Just consider what it fundamentally is: a company at the leading edge of a product category that has found absurdly strong technology/use-case fit, and is growing insanely fast.
Looking for a moat in the technology is always a bit of a trap - it’s in the traction, the brand awareness, the user data etc.
> Looking for a moat in the technology is always a bit of a trap - it’s in the traction, the brand awareness, the user data etc.
Traction, brand awareness, and user data do not favor Windsurf over GitHub Copilot. The few of us who follow all the new developments are aware that Windsurf has been roughly leading the pack in terms of capabilities, but do not underestimate the power of being bundled into both VS Code and GitHub by default. Everyone else is an upstart by comparison and needs some form of edge to make up for it, and without a moat it will be very hard for them to maintain their edge long enough to beat GitHub's dominance.
Definitely take that point. But this valuation is perhaps more about how much that traction, brand and data is worth to OpenAI, who cannot buy Copilot. $3bn doesn’t seem so disproportionate in that context especially given the amount of money being attracted to the space.
Define losing? My company pays for Copilot but not for Cursor, and it's not at all clear to me that we're the exception rather than the norm. What numbers and data are you working with?
Incumbent advantage of being in VS Code already? Thing is, Cursor is basically just VS Code, there's hardly any barrier to switching, so it's quite a weak advantage.
That's not actually how unseating an incumbent works. The incumbent can adapt to the threat for quite a while if they act on it, they just have to not be Blockbuster. Copilot is showing every sign of making up ground feature-wise, which is bad news for the runners up.
In brand velocity maybe, but copilot is rapidly reaching feature parity with cursor and will invariably overtake it—while costing less to users.
Same with Google vs OpenAI. I tend to agree with the sentiment that I most frequently hear which is that OpenAI is the currently popular brand, but that can only carry them so far against what will eventually be a better offering for cheaper.
Yeah it seems like there's really no "adult supervision" at all in OpenAI. This purchase was a panic move. Windsurf would be worthless without the AI. Probably OpenAI knows that AI is now a commodity technology and no longer a space they can monopolize so they're just trying to get off a ship that's sinking, and find some viable path to having a tech that doesn't ultimately depend on OpenAI even having a monopoly any longer.
Cursor purports $300M in annual recurring revenue (ARR) but stays silent on churn.
They made $25M from subscriptions one month, took that number, multiplied it by 12, arrived at $300M and everyone has been running with that line without ever asking what their churn looks like.
They could have churned $24M the next month, ask yourself why they are silent on churn if they are doing so well.
Venture capitalists aren't ignorant, their business revolves around knowing exactly what churn is. Cursor has raised $1 billion with a $9 billion valuation. VC's willing to put in that much money has looked at their data and knows what the retention rate is.
If their plan is to make their money back selling the company, then they don't care about revenue or retention rate. The company just need to look like it might be doing well.
No, venture capitalists aren't ignorant, but their goal also might not be to build and run a healthy company long term. It might be to turn a quick profit by selling a startup to another company.
If that's so, then why is Codex such an inferior product to Claude Code? And why haven't they already built an code editor or at least VS Code extension yet?
JetBrains has been making IDE for a decade. They were the only company that actually made money by selling IDE. So I assume they have the best programmers who understand IDE.
However they fail to make a Cursor competitor so far. This alone suggests it's a harder task than meets the eye.
It is, but you are assuming that only a well known IDE team would do it. To me JetBrains is the least likely to be an innovator here because they depend on their reputation for being a mature technology.
Someone like me isn't known at all but it means I have been able to experiment for a long time without pressure, which is how you do real innovation.
JetBrains as a company probably owns 10 million lines of code and it's just really hard to move fast when you're tugging that kind of ball and chain
That AI is in a bubble akin to the crypto craze from a few years ago, and the valuation of these companies is divorced from their underlying business fundamentals
Steel man: Windsurf own the customer relationship. The models are just generic interchangeable services they use for processing.
Realistically: I don’t know how many users windsurf actually has and I never actually met anyone that uses them. Whereas Cursor AI took a huge percentage of the VS code users I know in real life.
If OpenAI just provides AI, then the various IDEs development wrappers / IDEs / low-code etc. can collectively bargain against OpenAI for low rates. If OpenAI has an alternative, then they can charge higher rates for all plugins/ etc. and give the market an alternative.
Enterprises won't require less software. If they require fewer software engineers, that would be those few engineers producing so much more software with better tools, for example, AI wrappers.
Yeah but we can already see that it doesn't work like that.
If you need to write a lot of code I guess, but that's really rare, like saying "I need to write a lot of laws. I need to write 50 new laws by Tuesday with at least 15000 words of new regulation to one-up my rival legislator who wrote 40 new laws last week"
If software engineers are more effective, I would expect there to be more software engineers. They’ll put out more and better code. More code means more engineers.
The contrary view is like saying gold miners are finding more gold, and it’s easier than ever, so we expect folks are going to leave town.
That's fair, but enterprises are often naive and prone to groupthink.
It was just a few years ago when automakers and rental car companies unanimously decided (has they had been told to decide) that COVID-19 would reduce demand for cars. They cut production, sold off fleets, and almost immediately found themselves unable to keep up with demand.
LLMs aren't the cause of mass layoffs and hiring freezes. The end of ZIRP, uncertainty in the macro and offshoring are the cause. AI is just something executives like to say recently when they do layoffs ("we'll be more efficient with cutting edge technology!").
I think there's real pressure from investors to show that some of your human costs will be going away.
After all most of those investors are deeply invested in AI technology already. At the valuation, they need to be able to show that it replaces human workers because that's the specific kind of greed that is driving the value of the stock.
And if you see your competition tighten their belt then you should tighten yours right? So without proof companies are acting like they can use a small number of human-ai hybrid workers. There's strong peer pressure to think that way as a direct result of AI
You can't really make a living anymore being a furniture maker. Even semi-famous people in the industry have a hard time doing it. Only a few make enough to feed their family.
Making cabinets, etc.. sure. But woodworking has drastically changed, and maybe programming is changing that way, too.
Of course any tool that makes a job easier means that less people are needed to do the same job. If demand doesn’t change the supply will naturally shrink.
So hypothetically 1 man can cut wood but it takes him 2 days to do a big job. With a power saw it takes him half a day so his output on this section of the job is amplified by 4x. Any tool that makes his life more trivial increases his output and therefore increases the supply of the product without touching demand. With an over supply the system will naturally lower in supply by replacing carpenters.
This happens for anything and any tool that makes someone’s occupation easier. You have to think in aggregate. It may be the increase is imperceptible as it only increases the efficiency of a worker by 1 percent which is nothing but in aggregate that translates to a 1 percent reduction in the work force. Of course reality is more complicated than that but I hope the example shows you what I’m saying.
And it gets even more complicated than this too because increasing supply can also increase demand because the product becomes cheaper. Or demand may have already been astronomically high so the increase in supply only meets the demand.
In general if the product is in equilibrium of supply and demand and you increase the efficiency of the worker producing the supply then you will reduce worker population because the job doesn’t pay well enough anymore and people leave or less people join. The system slowly comes back to equilibrium or it can oscillate back and forth between over supply and undersupply as it’s basically a control system. This is what’s been happening with software for the past 3 decades.
The idea that the power saw didn’t replace a carpenter is flat out wrong. The story is much more complicated than that but the reality is that in general it did replace some carpenters just like how vibe coding for sure is replacing some software engineers.
It's about popularity. OpenAI lost their monopoly now that there are many competitors so they're just trying to make a move to purchase "relevance". They're just trying to buy their way into the cool kids club, to remain relevant to at least a large number of kids.
Windsurf and Cursor feel like temporary stopgaps, products of a narrow window in time before the landscape shifts again.
Microsoft has clearly taken notice. They're already starting to lock down the upstream VSCode codebase, as seen with recent changes to the C/C++ extension [0]. It's not hard to imagine that future features like TypeScript 7.0 might be limited or even withheld from forks entirely. At the same time, Microsoft will likely replicate Windsurf and Cursor's features within a year. And deliver them with far greater stability and polish.
Both Windsurf and Cursor are riddled with bugs that don't exist upstream, _especially_ in their AI assistant features beyond the VSCode core. Context management which is supposed to be the core featured added is itself incredibly poorly implemented [1].
Ultimately, the future isn't about a smarter editor, it's about a smarter teammate. Tools like GitHub Copilot or future agents will handle entire engineering tickets: generating PRs with tests, taking feedback, and iterating like a real collaborator.
The thing is: we should not need standalone editors just to use AI coding agents. They could be just plugins, but Microsoft does not want to bend the plugin API enough for that. Windsurf has a "plugin edition" for JetBrains IDEs that works really, really well[0] (they also have a VSCode plugin[1] but it's lacking in comparison).
However, given that JetBrains also have their own AI offering[2], I'm not sure how long that will last too...
I am constantly surprised how seldom aider is mentioned in threads like this. I understand that it's not directly integrated into the editor, but the "editor + parallel CLI tool chain" paradigm feels so natural to me because we drop to terminal for so many other parts of building software. If you haven't tried it (particularly the architect/editor modality), it's worth a couple of hours of experimenting.
Aider doesn’t provide any interface that’s integrated into the editor tool, as you point out. That might be true for other similar side-by-side tools that I am not aware of.
But, if you tell aider to watch your files, you can drop a specially formatted comment into your file, and aider will see that and use it as a prompt.
So the integration is sort of “implicit”. Which sounds kinda like the cheap way to go, in comparison to the current brand name tools that have their own chat boxes, dropdowns with mode selectors (ask, edit, agent), and so on.
But look further into the future and an ambient interface is probably where we end up. Something where the Ai agent is just watching what you do, maybe even watching your eyes and seeing what you’re attending to, and then harmonizing its attention to what you are attending to.
Tight editor integration means better diffs (right in your editor), easier context manager, and other convenience features that CLI-only tools can't have.
This doesn't mean that aider, claude code, etc. aren't very good tools, but it does make sense to distinguish between built-in tools vs external ones. A similar non-AI example is debugging or linting: IDE integration makes it much easier than using a separate tool.
Our plan is to be a superset of Cline and Roo's features (we already have all the major features from both) [0]
We also have our own provider, which means no need to bring your own API keys (you can if you like, but it is batteries included by default) and we're not charging anything on top of the API pricing. Instead of monetizing on individual developers, we want it to be free for them and make money eventually off enterprise contracts [1]
maybe you could answer a question about kilo usage: If I choose Google Gemini as the API provider and give it my Gemini API key, why does it say that I'm low on credits (and I get API request failures immediately)? As far as I understand gemini 2.5 pro preview is free to use. (and in Cline I'm able to choose Google Gemini as the API provider & provide my API key and it will successfully make API requests)
Yes - with our built-in provider, we provide all the models that OpenRouter provides but without OpenRouter's 5% markup. We provide them at cost (the AI provider cost)
I suspect JetBrains will never limit this. I've yet to recall anything in the past where they have done this even when they have a similar offering.
In fact, their own AI extension appears to be pluggable in and of itself. I think they see the value in being easy to adapt different AI solutions to rather than trying to only provide their own.
JetBrain's main business model depends on buying the editor, and if users still see the overall editor better, any AI plugin support will likely just increase the sales.
Microsoft has been dragging their feet when it comes to updating the LSP spec. Many of their Copilot features are done in VSCode, in fact using private APIs that are not accessible to other extensions.
I am all for everyone adopting LSP, but the reality is harsh.
LSP is amazing but also kinda sucks balls. It’s impossible to run VSCode without a million pops in the corner with a million extension errors. It’s so bad.
And autocomplete is the least interesting thing an LLM can do. Cursor’s UX isn’t the end game but has lots great features.
The ideal UX is still being worked out. It’s good that different people are building tools to try different ideas.
Is windsurf essentially the same as cursor? I didn’t realize there was something similar for JetBrains but if it’s a cursor-equivalent for JB that would be wonderful.
If Microsoft were smart, they'd just acquire Cline (or fork it), make it an official VSCode feature and be done with it. It smokes Cursor and Windsurf and it's a free plugin you can just install in un-forked VSCode.
> At the same time, Microsoft will likely replicate Windsurf and Cursor's features within a year. And deliver them with far greater stability and polish.
I agree with the first part, I'm much less optimistic about the second part. I suspect they will create something that is worse, but cheaper if you already pay for Github/Office 365/whatever. Then many large enterprises will switch to save money whilst the engineers complain, just like with Teams.
They already succeedd well enough that VSCode is the only Electron app I tolerate on my private systems, naturally on device assigned ones I have less control.
I mean they already have. GitHub Copilot was the first LLM coding tool before "LLM" was in the lexicon. MS/Github kind of squandered their lead with it, but they released Agent Mode a few months back https://github.blog/news-insights/product-news/github-copilo...
I use vscode for personal javascript projects but the time I spent on a .NET team using VS was an incredible downgrade compared to years and years of intellij. I ended up leaving because tech debt/bugs kept causing weekly overnight on call incidents that we were never given time to fix, but when they asked who wanted a Rider license I got myself on the list immediately.
Copilot owns the platform, had an amazing head start and yet still is the worst option available. I don’t mean to be harsh but this was a titanic fumble.
GitHub has been failing upward for more than 5 years. They could have totally dominated software development and security - failed. Could have been the undisputed champion of code hosting - failed. Should have dominated development co-pilots - failed.
I actually find it a little reassuring that they can't seem to get out of their own way.
It's a close call - I make this based on the fact that GitHub is viewed as an anti-choice by some in the community, a huge change from the "you don't use GitHub?!?!" energy they had pre-acquisition.
The MS acquisition traded the developer community to briefly appeal to enterprises, then quickly let both down.
Both the startups I worked at and the mega corps are all on github or moving there from bitbucket. They are in a bit of autpilot mode in terms of useful new features aside from actions but I can't think of any new bitbucket feature since I graduated and started working.
Bitbucket is not a player, as you said there are only people leaving. Gitlab has a better enterprise posture than GitHub and can be deployed more securely. Most developers aren't unhappy with GitHub, but IT and security teams are.
To be fair, they have been behind the competition for many years. Gitlab had extremely good CI, security scanning, organisational concepts, etc. for years before GitHub introduced their ones (and Actions still has a worse UX, and GitHub still doesn't have anything below an organisation).
GitLab UI is inferior IMO, and I've used both quite extensively.
I don't like that GitLab lets you nest organizations and such, it makes it so painful to find things over time. I appreciate GitHub doesn't do this, I view it as a plus
I also disagree about GitLab CI, not that it wasn't smart for them to include alot sooner than GitHub, but Actions is really good and really easy to get up and moving with. I find they run faster, have better features - like they can annotate a PR with lint errors and test failures - with very little comparative configuration.
GitLab CI yaml is a mess by comparison. GitHub was smart to push things to the runner level once a certain complexity threshold is hit.
This has been my experience of course, and so much of it is really subjective admittedly, but I don't think GitLab is truly ahead at this point.
> I don't like that GitLab lets you nest organizations and such, it makes it so painful to find things over time. I appreciate GitHub doesn't do this, I view it as a plus
Nah, I hate that. At my job we have a few different orgs, with terrible SSO boundaries (having to auth multiple times to GitHub because I work on repositories from different GitHub orgs). Allowing you to have a proper structure with nestedness, while still having good search, is great. You can also easily move projects and namespaces around, so if the structure doesn't work, it can evolve.
Why would you have the 50 library repositories you've had to fork as top level projects polluting your org? You also can't really do shared variable, environment, CI configs between repos of the same project/type.
And it being open core (MIT) means spinning up a version to test something is incredibly easy. Not exactly resource cheap, as it's still a rails app with multiple servers "smuggled" in the docker image, but it is easy
And I have long held that they are hungry, shipping like clockwork on or about the 20th of every month, showing up with actual improvements all the time https://about.gitlab.com/releases/ It seems this month brings 18.0 with it, for whatever that version bump happens to include
They also have a pretty good track record of "liberating" some premium features into the MIT side of things; I think it's luck of the draw, but it's not zero and it doesn't seem to be tied to any underhanded reason that I can spot
Yeah, it's almost certainly the network effect. Although poor GitLab isn't doing themselves any favors by picking what seems to be the slowest web framework one can possibly imagine
But, anytime I am empowered to pick, I'm going to pick GitLab 100% of the time because it has every feature that I care about and "being popular" isn't a feature that I care about
Well you’re right (especially wrt things like security scanning), but you sort of have to include Azure DevOps in the conversation nowadays. I think the end goal for Microsoft is to get the larger organizations into ADO, either cross-pollinate pipelines and actions or just replace actions with pipelines at some point, and leave GitHub for simpler project structures and public codebases.
That’s why you won’t see a ton of work go into e.g. issues/projects on GitHub. Those features all already exist and are very robust in ADO, so if you need those kinds of things (and the reporting an enterprise would want to be able to run on that data), then you belong on ADO.
I can say with a high level of confidence that the goal is definitely not to push larger orgs to ADO over GitHub. ADO is and will continue to be supported and you’re right that its project management features are much more advanced than GitHub, but the mission is not to push people off of ADO and into GitHub.
I'm curious what the cost per user is on Copilot. It doesn't make sense for them to be a loss leader so they're probably running the model at cost or a profit compared to the startups that have more of an incentive to scramble for market share.
I'm too lazy to grab my work laptop, but one of the funniest things about copilot to me is which one? There's M365 copilot, Teams Premium (which gives you copilot in Teams), browser extension, the coding plugin, and others. It's been extremely time consuming to field requests from our users because every time our help desk gets a request for it, they have to have a conversation about which one the user is asking about. They don't even know, and of course I can't blame them.
For someone that never used windsurf, what features does it have that GitHub copilot does not? Reading their webpages I didn't spot any "killer feature" that would convince me to switch.
I always felt that cursor and windsurf should be just extension to vscode instead of a fork. Was there some missing functionality is vscode that was missing? Is it still missing?
There are some extensions that work in this way and allow to use multiple implementations depending on task at hand without any long term commitment.
I feel like such fragmentation is by artificial just to lock users in single ecosystem.
It can write a lot of code, that works, better than vscode can (right now).
It's in a lot of ways the OpenAI story itself: Can they keep an edge? Or is there at least something that will keep people from just switching product?
Who knows. People have opinions, of course. OpenAIs opinion (which should reasonably count for something, them being the current AI-as-a-product leader) is worth $3B as of today.
I doubt they'll restrict it to their own models. The amount of business intel they'd get on the coding performance of competing models would be invaluable.
The differentiator of Cursor is it’s way smarter at basic code completion than GitHub Copilot. I pay for Cursor instead of GitHub Copilot even though I get the latter for free from open source contributions, and I made that decision after five minutes of usage after using Copilot for what, more than a year? I won’t even talk about how Cursor guesses where I’m going to edit next and makes the correct edit most of the time, just the fact that Copliot makes completions that result in unbalanced parentheses/braces all the time and Cursor doesn’t makes the switch a no-brainer; that’s not even a fucking AI problem, you just need to look around and see that function you just completed already has a closing curly brace, all it takes is some traditional AST analysis if your model is dumb. (Copilot made zero progress on that issue during my time using it, but I can’t say if that was fixed after I ditched it.)
Same. Cursor might be the only tool I've purchased a year's subscription to before the end of my free trial.
I've tried just about every model on its own over the years, and yet there's something about the Cursor workflow that frequently still gives me chills when it shows me again that it had clearly anticipated what I would think next in a way I just don't experience with other tools.
Holistic seems like the right word?
If it's all smoke and mirrors as some folks imply, then it's Penn and Teller level smoke and mirrors. Beware those who tell you that they could duplicate anything of value in a weekend.
Windsurf does it all the time like a wife of 40yrs completing your sentences. A good example is when you remove a function parameter. It automatically prompts to remove the arguments in all usages of the function, saving me a lot of time.
It's better at coding, but they are essentially paying for users.
I would also argue that the product could be built over two weekends with a small team. They offer some groundbreaking solutions, but since we know that they work and how, it's easy to replicate them. That also means they have significant talent there.
Hence, they are also buying the employees.
The code base itself is basically worth nothing, in my opinion.
I'm going back and forth between Windsurf and Github Copilot right now. Windsurf's development iteration speed is much fast and features are added faster.
For example, Github only autocompletes based on what file you have opened in the current editor's tab. Windsurf indexes your entire code base and seems able to autocomplete based on what other files you have in your project. Autocomplete also spans across multiple lines and open tabs.
Windsurf's agentic tool (Cascade) can run terminal commands and read the output without opening a terminal like copilot. It can undo the agent's actions easier than Copilot. Though I think Cursor is superior in that regard, it can undo multiple checkpoints.
Still evaluating Windsurf but it, Cursor, and Claude Code are all more sophisticated than Github copilot at the moment. I'm sure copilot will catchup but by that time the other tools may have already iterated ahead.
If a company can align it's business model with user goals, then it can work in the long run. Apple has somewhat aligned it's integrated hardware sales business model with user privacy. Google and Meta are advertising companies and capturing user data and attention will always drive the business.
Yes, but it's not a meaningful part of their revenue unlike Google where it's' their entire revenue.
They are very different companies in structure and it certainly is a "pick your poison" but it's completely stupid to act like they're the same on this front. Apple is better on user privacy
...unless you care about state actors, which you should, in which case your data is the US government's either way.
Satya's talked about how some acquired companies such as LinkedIn and Github are allowed to operate independently for the most part and keep their culture. Or else we'd all be using Teams instead of the LinkedIn messaging feature!
Do you consider the Microsoft-managed plug-in marketplace and infrastructure to be a private or public resource? From my understanding Microsoft has never been vague on the position that the plugin marketplace is exclusive to the official VS Code distribution, and the TOS specifically forbids forks from doing so.
Cursor and other forks have decided to circumvent this, some even going so far as to use proxies to bypass restrictions.
I'm not convinced Microsoft owes other billion dollar companies free access to a product they've built, curated, and supported for over a decade. Plug-in authors are not restricted from publishing their products on competing marketplaces.
Microsoft is owing its bad reputation to Windows, Office, Sharepoint!!!, Teams (and more?). The quality of developer tools and languages (C#, Visual Studio, Code and .NET Ecosystem, Azure UI is also great) from Microsoft has been flawless (with some exceptions like webforms, or ui code generation tools of the past).
I wouldn’t say that. JetBrains is incredibly bloated and has significantly less community support.
I’ll agree on Teams being crap though, mostly for how dumb it is that they’ve rewritten it multiple times and created a confusing slate of weird versions like “Teams (work or school)”
Their devtools team is surprisingly competent when they choose to be. Pre-2015, people used Sublime Text, Atom, Textmate, Notepad++, Light Table, Brackets, Emacs/Vim, Intellij. VS code single handedly crushed all of them with code completion and language servers that require zero configuration. Emacs/Vim lost share, Jetbrains (and also Eclipse) were forced to release their own "lightweight" code editors, and everybody else became mostly irrelevant (except perhaps Sublime Text since it has the best native performance out of all editors).
100%, I swore by emacs, but then switched to vs-code recently, and believe-me, switching editors is one of the hardest things to do due to ingrained muscle-memory - but vs-code made it easy with emacs-mode etc.
vs-code is one of the few products coming of of microsoft that leads the pack by a big margin, and it is no surprise that all of these startups are forking it.
I would contend that JetBrains has only grown even with VS Code around. They're still more than viable, support things on a near similar cadence (and even in some cases, faster and/or better) than VS Code gets support for it.
I agree with the rest, they've all mostly lost market share or completely no longer exist due to VS Code, but not IntelliJ, that platform is going really strong.
Though no doubt, VS Code has pushed JetBrains to rethink some things, and be better in general.
Free doesn't matter here. JetBrains is an established toolset that people pay for. They've already been competing with free, and free didn't put them out of business. In some ways, free likely made business better than ever (I know alot of devs that started with VS Code and moved to JetBrains for various reasons)
They can have all the money in the world and it doesn't mean much in this context.
For while Microsoft is going to invest heavily in a Cursor / Windsurf like product and likely do alot to ship it in their editors - likely with exclusions or lag times between updates on other platforms - there's zero reason for Google to do this for example, when they could sell through Gemini for Code as an extension across all editors.
I don't see JetBrains having issues because of AI tooling, for most of these companies, its a boon to be on the JetBrains platform. Especially because JetBrains has lots of enterprise customers who would naturally be very interested in buying AI tooling for their developers. Its a natural market
I was a little late to jump on the cursor bandwagon but finally downloaded it because i liked the LLM chat interface in the sidebar. By the time my free trial ran out, VSCode had added a LLM chat interface in the sidebar. Yes Cursor had a bit better auto complete and maybe a few other things but it wasnt good enough that it was worth paying for.
But I'm glad OpenAI is getting into the tooling space in this way. I cant wait to use all the cool features they build after VSCode rips them off.
> By the time my free trial ran out, VSCode had added a LLM chat interface in the sidebar.
I am guessing you are talking about GitHub Copilot when you say VSCode. GitHub Copilot is far far inferior product when compared to Cursor, Windsurf or Augment Code. Most people who try almost any Copilot alternative for a reasonable amount of time end up canceling their Copilot subscription. I did, after two months of using both.
How long ago was that? 6 months ago I switched from VS Code to Cursor, which at the time was FAR superior to Copilot. Around a month ago I switched back to VS Code, and found there's not much difference any more. Autocomplete in VS Code is still less good, but the agent mode in VS Code feels pretty similar to Cursor's (albeit a little slower, perhaps).
To be honest I think both are quite limited by context length (in that they try to limit the context they send to the LLM and hence cost), and so I find myself using Gemini 2.5 in AI studio with the 1m context length, and asking it to generate instructions for Copilot (which seems to work pretty well)
I mean, the fact that OpenAI, at the bleeding edge of it all, has decided to buy an IDE is a rather strong hint that the future of agents handling entire engineering tickets might be further out than many believe.
If autonomous agents were just around the corner, then why wouldn't OpenAI bet on their own Codex product obviating (most) need for an IDE and save themselves the $3 billion?
These products are not complicated at their core — you can pretty much just drop in something like Monacopilot [1] and be 80% of the way there. But the last 20% is a real slog, and it mostly comes down to handling edge cases (bracket closing...) and optimizing prompting/context so you aren't burning cash. Whatever anyone claims about "feeling the AGI," AI isn't there yet.
They did. They’ve just released codex (CLI client).
They don’t have access to copilot users in general, Microsoft and Google does. And perhaps they are realizing that Microsoft is hedging them over multiple LLM providers and maybe no longer feeding them juicy copilot data, with humans in a tight loop, correcting LLMs.
This is a good point. It is already the case that unless you deeply review every Windsurf change you will have zero understanding of your codebase. If it gets 1000X better in the next 3 years why would anyone look at code at all?
Of course, back to reality. Today, at least in my workflow, I use / like Windsurf but it is a small part of what I am doing. For any code I want to keep I mostly write it by hand (using vim for a very bare-bones / cognitive mode experience). For me, the real flow state occurs in vim while ChatGPT and Windsurf are great for exploration.
It sounds like the openAI team is overburdened (I guess they aren’t AI super users yet) so this may be their only option. Easy entry into a key segment, at least for now, and locks out competitors.
As a competitor in that key segment I don't feel locked out. I could almost jump for joy that this very weak-tea move is the most they can do with that much money. They're just quintupling down on the technology of 50 years ago. There's no threat to me at all here as a creator of from-first-principles IDE technology.
Yeah. Every time I see entirely unfounded claims like that, I remember that I've been seeing them for literal years now. While there have definitely been improvements in AI capability, they have largely been very marginal, while the claimed "will handle entire engineering tickets" capability requires huge leaps in capability and reliability that _we just have not seen evidence for._
Mentally, I'm replacing claims like this with "it will do magic!" and I think I'm just about as likely to be correct.
It wasn't any better for me. It deleted all my code. The answers were like it was a completely different model. I used Windsurf once and never opened it again
>Ultimately, the future isn't about a smarter editor, it's about a smarter teammate. Tools like GitHub Copilot or future agents will handle entire engineering tickets: generating PRs with tests, taking feedback, and iterating like a real collaborator.
I think a few options for this already exist, but honestly they don't go far enough. I want something like an AI scrum master, for hyper agile teams, that can task out smaller tickets to AI sub agents.
I would integrate this thing in with something like an AI powered Jira.
Two arguments exists.
1. I need to take about 6 months off and start building this now, even if I don't know exactly how I'll get it done. Between a combination of vibe coding and maybe a bit of outsourced work ( looking at Eastern Europe), I could get this done with my personal funds.
2. To do this properly would probably require tens of millions of dollars. I'll probably burn myself out trying to do it solo without ultimately getting to a sellable product.
The biggest issue here is to actually scale I would need to either have users bring their own LLM keys or have tens of thousands to spend on LLM tokens.
I view this as an another step in the push/pull between local things, remote things, local things remotely, thin clients, network partitioning, cloud, zero trust, etc...
The last cycle I remember of this IMO is iPython -> Jupyterhub/Jupyterlab. Of course, iPython has existed for a long time, though that change was made because data was too big to analyze locally and it turns out it was more convenient to centrally manage kernels/images/libraries for convenience.
MCP servers and Cursor/Windsurf changed that a bit, but it will end up centralized again at some point (or at least aggregated, if it's not already?). People are passing around lists of interesting MCP servers now, and that will be out of fashion in less than 12 months.
> At the same time, Microsoft will likely replicate Windsurf and Cursor's features within a year. And deliver them with far greater stability and polish.
Microsoft software quality has gone downhill recently, and I'm not going to bet on them delivering something more polished than WS and Cursor here.
Side: all images on Microsoft websites are low resolution! it's like they don't even check their own website.
Their “programmers” are more busy with making blogs and videos than functioning tests or technical documentation, and they start using JavaScript and Python for everything.
I’m not surprised their quality went to shit. There are some pearls left, C# in general is pretty good, and Aspire is becoming quite neat.
The latter I think mainly because David Fowler is just a great developer
Because if you're referencing to a headline (without reading the article) that was on H a couple of days ago, it stated that 20-30% of the code in the repos was written by software.
Software != AI
To quote wongarsu in the same post:
"Considering that most of their software has been developed for decades and AI assistants have only started becoming useful in the last ~4 years it would be very surprising if 30% of their code is AI written. I doubt they even touched 30% of their code in the last 4 years. But what is perfectly plausible is that 30% of their code is written by code generators. Microsoft has a lot of interface code. All the windows DLLs that are just thin syscall interfaces, the COM and OLE interfaces in their office suite and everywhere else, whatever Office uses nowadays for interoperability to allow you to embed content of one product in another, whatever APIs their online products use, etc. In the leaked Windows XP source code it can be difficult to find the actual source code in between the boilerplate files containing repeated definitions, and in the decades since then the world has only leaned more into code generation."
> Ultimately, the future isn't about a smarter editor, it's about a smarter teammate. Tools like GitHub Copilot or future agents will handle entire engineering tickets: generating PRs with tests, taking feedback, and iterating like a real collaborator.
I disagree, but would love to be wrong. These tools exploded onto the scene and were massive productivity helpers, but since their initial integrations they’ve churned rather than improved in the last 2 years. They are even worse when you try to iterate rather than just get them to one shot the problem space.
> At the same time, Microsoft will likely replicate Windsurf and Cursor's features within a year. And deliver them with far greater stability and polish.
At the speed that AI programming is going, there will be something else that they are falling behind of that will exist in a year. Just like Agents now, they are adding them, but will always be a step behind progress.
This is the right take, but long term. Short term, it's just about investor hype. Cursor is becoming more mainstream and if OpenAI falls behind on this, they'll be losing momentum. But yes, the fields moves so fast, it'll be totally different in a year or 2. Does anyone recall langchain?
Non-public numbers may as well be pulled out of thin air. WeWork was a $50bn company according to its VC bagholders, and that was marked down by 80% once they released their books to the general public.
I never did like JetBrains primary product, IntelliJ. It felt clunky even compared to Eclipse for Java, let alone VSCode for … everything. DataGrip is the lone standout imo, but as of the last update I paid for, it didn’t have even basic copilot
I use Cursor in anger every day. The core idea behind Cursor is genuinely smart. But the execution is like the classic "unfinished horse" meme [0].
Microsoft provides the editor base, foundation models provide the smarts, and Cursor provides some, in my experience, extremely buggy context management features. There is no moat.
I have. It’s ~fine. The only feature difference it has currently to vs code that makes a difference to me is allowing multiple files for rules.
Meanwhile GitHub web integration is approaching seamless in vs code. To the point I often forget I’m in a browser instead of the app, until an extension I use doesn’t work.
I've tried both Cursor and VS Code with AI in the agent/edit mode. They both seem similar enough. Is there another mode I haven't found where Cursor has a distinct advantage? If so, I'd like to try it.
I gave up on Cursor because my trial ran out, while VS Code with Copilot doesn't seem to charge me anything.
I have tried (stopped a couple months ago). The Python extensions broke all the time while they manually patched around the latest MS release a few days later. Syntax highlighting glitched every other day requiring a full reload. Remote dev via SSH or tunnels also randomly stopped working. Liveshare...
Essentially they do not own the platform their core product is built on.
Maybe it's fine if you only do local development in other languages (Javascript?), but I completely swore it off.
As a counterpoint, I also use cursor as my daily driver and I have been tempted to switch many times because of the endless bugs. Just take a look at their forum.
yeah Microsoft could never conceivably develop an extensible source available IDE people love so much they even fork to build $3B companies on the scraps of. absolutely alien!
I'm frankly very skeptical of your last paragraph. That's not at all what seems useful to me. But we'll see!
But I agree with you about the first part, and I think it's awesome for me as a user that all this competition to build a matter mousetrap is happening right now! I'm not as certain as you are that Microsoft will end up building a better version. It's definitely one of the likely outcomes. But it's also totally plausible that Cursor or Windsurf can win the race, even if they need to replace every single one of the MS extensions and entirely diverge the core IDE from upstream. These products are well capitalized and it's just not that hard to build the core pieces of an IDE.
UI may be close. Functionality is very very different. Copilot is $10/month. Cursor is $20/month. I canceled my Copilot subscription after 2 months of using both. Compares to competition, Copilot has been garbage for quite some time.
Why should they have restricted the marketplace? It's really annoying imo that they lock vs codium out of the more useful plugins like the SSH remote one. However luckily most only take a setting or two to enable anyway.
I am slightly more optimistic, because the API may not be fully centralized- there may be more than one foundational AI company in the end. Like WhatsApp exists because there's the iOS/Android duopoly, an agent-neutral IDE from a non-foundational company without its own API aspirations may continue to exist
Microsoft is slow af for a company that size. Maybe yeah, they are slow because of that size. Don’t bet on them out accelerating a startup, the evidence so far in the past year is that they will stay a year behind every year
The sheer number of AI written message board posts might just make me stop reading the comments on sites like Reddit and HN. I wanted to stop anyway, this seems like a good push to encourage me to wean myself.
Every one of us leaving (not engaging/commenting) increases the share of AI generated comments (vs real users) the next iteration will train on. I'm not even sure which option is worse. Withdraw and let everyone dilute their own training data, or stay and feed them our mindset and experience...?
To be fair, only 3 posts within "possible LLM usage" timeframe. Also I don't think using LLM to comment == bot.
More curious about the motivation behaviour such behaviour, if it is occurring
This is beginnings of AI discrimination. If an answer is written by an LLM but equal or superior in quality to a human answer why question or disparage it?
I don't know but it looks like you're probably a white guy. Your mannerisms and vibes make it look like you're white. Nothing wrong with this, just wanted to point it out. See what I'm saying.
I find this comment odd but not at all discriminatory and will happily inform that yes, I am white British.
I can see the point you’re going for but I do think it’s more complex than that. I don’t think LLMs are equivalent to another human race.
Principally, LLMs have many pretty major differences to humans which you don’t really see at the inter-racial level.
Secondly the reasoning for why asking someone their race in this context would be weird involved a lot of human history. If you’d asked my height or eye colour there’s nothing discriminatory feeling about that. That historical context doesn’t exist with LLMs
The LLM provided an answer that has superior quality or equal quality to a human. Then instead of judging the statement rationally on this quality we decided to judge the question on whether or not it was AI.
This is the same irrationality we used to discriminate humans. There is no difference in logic. The reasoning you used here about how LLMs are not equivalent to human beings is the same reasoning Hitler used on Jewish people.
And here’s the thing. I agree with you. If you gassed and holocausted LLMs wouldn’t give two shits.
The main point here is that the logic and irrationality and evil present in racism is all at work here. We are literally being discriminatory, there’s no difference. Everyone missed the point about the quality of the statement itself and immediately based their judgement on whether or not it’s AI as if that was actually a rational thing to do. (It’s not).
> If they start walling off features like TypeScript 7.0 from forks, the open source pushback will be fierce—and that could backfire hard.
Do they have the man power to compete with Microsoft?
Linux managed to do it but Linux is the biggest, most successful free software project there is. Firefox and its forks are a better example. If Mozilla stopped working on Firefox, the forks would be pretty much dead in the water: they simply do not have the man power necessary to maintain a modern browser.
Does microsoft have the wisdom to predict where this line of technology is headed, and/or the agility to course correct when their predictions don't quite hit the mark?
Cursor blows copilot out of the water in my experience. Man power clearly isn't the most decisive factor in this battle.
Copilot is limited to 64k context window. Even if the underlying model is gemini with 20x that. It’s gotta be a major reason copilot is so bad in comparison. They are all the same sets of models under the hood
> Do they have the man power to compete with Microsoft?
There are a fair number of examples where smaller companies and/or open source beat Microsoft's entrenched products. Usually a key indicator is that MS's products stagnate (which doesn't yet appear to be happening currently).
Microsoft’s army of cheap offshored labor isn’t going to be useful for something like that. And they already have copilot, which is miles behind cursor, where was the manpower on that?
- A $3B signal that OpenAI is unable to do product
- AI assisted coding is mostly about managing the context and knowing what to put in the context to avoid confusion and dumb mistakes, it's not about the UI.
- This signals that OpenAI believes that highly effective coding assistant LLMs will become a commodity / open source and so UI / tooling lock-in is a good investment.
True, but how long does it take to build something similar? I see it as a defensive move, probably good for the industry to let some people with innovative ideas in AI cash out now so they can do the next thing.
chatgpt is massively popular, I'm not sure that's the signal I'd get
they're acquiring one of the biggest the front doors to developers, with Windsurf - whether it'll _remain_ in fashion or not, that's a different debate. This can be like facebook acquiring instagram (if developers turn out to be the actual profit-driver niche for LLMs, which currently seems to be the case)
> developers turn out to be the actual profit-driver niche for LLMs
AI is definitely huge for anyone writing code, though one can imagine a model like o3 completely replacing 90% of white collar jobs that involve reading, writing and analysis.
Interestingly, o3 is particularly bad at legalese, likely not fully by accident. Of all professions whose professional organizations and regulatory capture create huge rents, the legal profession is the most ripe for disruption.
It's not uncommon for lawyers to bill $250 to $500 per hour for producing boilerplate language. Contracts reviewed or drawn up by lawyers never come with any guarantees either, so one does not learn until too late that the lawyer overlooked something important. Most lawyers have above average IQs and understand arcane things, but most of it is pretty basic at its core.
Lawyers, Pharmacists, many doctors, nearly all accountants, and most middle managers will be replaceable by AI agents.
Software engineers are still expected to produce novel outputs unlike those other fields, so there is still room for humans to pilot the machine for a while. And since most software is meant to be used by humans, soon software will need to be usable by AI agents, which will reduce a lot of UI to an MCP.
Your take on lawyers is absolutely insane. If you don't think the extremely specialized and well trained professionals can successfully navigate contracts then I can't wait for the absolute garbage the LLMs spit out when faced with similar challenges.
Honestly, same for doctors and accountants. Unless these model providers are willing to provide "guarantees" that they will compensate for damages faced as a result of their output.
Doctors and Lawyers are required in many areas to carry malpractice insurance. Good luck getting "hot new AI legal startup" to sign off on that.
While malpractice insurance exists for human docs and lawyers, there is not really any difference between an ai-powered lawyer drawing up a contract, an ai-powered doc reviewing a chart and recommending next steps, and a self-driving car making a turn.
The most obviously "lethal" case (cars) is already in large scale rollout worldwide.
At scale, self-driving car "errors" will fall under general liability insurance coverage, most likely. Firms will probably carry some insurance as well just in case.
LLMs already write better prose than 95% of humans and models like o3 reason better than 90% of humans on many tasks.
In both law and medicine there are many pre-existing safeguards that have been created to reduce error rates for human practitioners (checklists, text search tools (lexis nexis, uptodate, etc.), continuing education, etc.) which can be applied to AI professionals too.
Fair point, but how much billable legal work requires that caliber of skill? I'd argue that 80% of it could probably be done with an o3 or o4 caliber model with some safeguards built into the pipeline and perhaps a bit of specialized training or MoE guardrails, human review, etc.
I think the mistake people make is misunderstanding the slope of the S-curve and instead quibbling over the exact nature of the current reality. AI is moving very fast. A few years ago I'd have said that at most 25% of legal work could fall to AI.
Note that this massive change happened in less time than it takes to educate one class of law school grads!
openAI models are good at solid, fluent academic style prose. DeepSeek R1 can sound fresh and can use more "voices" that feel authentic to the reader. Grok-3 is close behind.
Writing good prose is a far different skill than coming up with a compelling and innovative plot and style.
As a data point, OpenAI now blocks o3 from doing the "continue where the story left off" test on works of fiction. It says "Sorry, I can't do that".
The next step for Cursor and Windsurf both is that they need to work together to provide an answer for what it means to be a VS Code fork in the new era where Microsoft is trying to strangle the forks. If they're not already they should be teaming up with each other and with the VSCodium team and with the Open VSX marketplace.
Microsoft is an existential threat to their model here, but with the money they each have coming in they together have the opportunity to make the whole ecosystem better by building out viable infrastructure for all VS Code forks, if they can cooperate.
There's a real opportunity here to build a sustainable, open ecosystem for AI-powered dev tools - but it's going to require actual coordination, not just parallel efforts
Investment vs. acquisition is going to have different price points.
At $40M ARR, I assume the founders don't really need to make more money and are not in a rush to sell. Therefore, the price would go even higher. This can't be compared with investment where the founders still retain the control.
Cursor is probably the fastest growing company in the history of our modern civilization. Achieving a really high multiple doesn't seem out of line.
I'm skeptical of Cursor but I can see why they achieve that high valuation.
If I recall correctly from the recent YC interview, the Windsurf founder noted their team leans more toward GTM than engineering. That makes this less likely to be a classic acquihire (as with Rockset) and more plausibly a data play rather than a product integration.
My current read is that this is a frontier lab acquiring large-scale training data—cheaply—from a community of “vibe coders”, instead of paying professional annotators. In that light, it feels more like a “you are the product” scenario, which likely won’t sit well with Windsurf’s paying customers.
Agreed. It seems like a data play and a hedge to beef up vibe code competition against upcoming Google and MS models so OpenAI doesn't lose API revenue. I would assume vibe coding consumes more tokens than most other text based API usage.
They have an old-school enterprise sales operation that is doing superb work. Apart from that, ChatGPT's projects are useless crap (can't read other convos in a project; can't generate project documents from a convo), and so clearly they would get value out of just getting some developers who have built anything of use to a poweruser.
Yup. Even a small market share is market share. Plus they are paying to acquire a team of folks who are already in this space and who will, until golden handcuffs come off, keep working in this space. Still an insane number though.
But openai is stronger brand with free publicity - whatever they say/do will instantly show up the same day on all news across the world.
The "space" exists for months, there are no people with 10y expertise here, with their brand they can attract any talent they can wish for in this "space", no?
You can probably vibe code 80% of it in a week or two?
I guess it's all up to interpretation, but having a brand in one space doesn't necessarily translate to a brand in another. OpenAI doesnt/didnt have a code editor. Now it does/it will.
I'm fairly into llms but it took me awhile to try cursor because the cost of changing editors is very high. I'd probably eventually try a OpenAI editor but only if I saw it was actually getting adoption and good feedback from others.
I'd also argue that while this llm powered editor space is pretty new, the editor space in general is much older.
OpenAI knows that everyday someone uses Gemini their ChatGPT brand dies a bit faster. Wonder what Google has in storage for I/O now in May, would be a death sentence to just steamroll with Gemini-3.
Open AI needed to spend $3B pivoting away from bigdata based AI. But instead they went for the most shorted sighted move possible of snapping up the "trendiest" company nobody has ever heard outside the Ycombinator echo chamber.
Typical VI-fallacy BS. If LLMs were actually good they would replace IDEs completely not be integrated.
There's an in between case, where LLMs are useful and give coders a (say) 20% speedup, and everyone has to use them. They don't have to be perfect to be a big industry!
But is there a secret sauce in any of the coding agents (Copilot Agent, Windsurf, Claude Code, Cursor, Cline, Aider, etc)? Sure, some have better user experience than others, but what if anything makes one "better at coding" than another?
As this great blog post lays bare ("The Emperor Has No Clothes", https://ampcode.com/how-to-build-an-agent), the core tech of a coding agent isn't anything magic - it's a set of LLM prompts plus a main loop running the calls to the LLM and executing the tool calls that the LLM wants to do. The tools are pretty standard like, search, read file, edit file, execute a bash command, etc. etc. Really all the power and complexity and "coding ability is in the LLM itself. Sure, it's a lot of work to make something polished that devs want to use - but is there any more to it than that?
So what is the differentiator here, other than user experience (for which I prefer the CLI tools, but to each their own)? $3B is a lot for something that sure doesn't seem to have any secret sauce tech or moat that I can see.
Yes and no. A messaging app with 450m users has very strong network effects. Users are sticky in a way they aren’t going to be with a VS Code fork which will be increasingly incompatible with the VS Code ecosystem. There are a lot of equally good alternatives to Windsurf and you don’t have to persuade all your friends and relatives to switch too.
The moat is Windsurf’s custom LLM and the ops around it (training pipelines, fine-tuning, infra).
Codeium (Windsurf’s parent) started as a GPU optimization company, so they have deep expertise there. Unlike most agents that might just wrap OpenAI/Claude/etc Windsurf’s own model powers its code edits, not external API calls.
That’s where the defensibility is. better in-house models + efficient infra = stronger long-term moat
You'd think with all these super hyper advanced AI tools they're shitting out they would be able to make a mediocre VSCode extension of their own instead of flushing 3B down the drain. Guess that's slightly out of reach of their "AGI"s though.
i fail to understand what makes this $3B valuation justified.
i built my personal code assistant after using cursor/windsurf/aider/cline because i was frustrated with how crappy they worked for my use case. i only program in python/js/html/css and i needed something better. only took me an hour of prompting and after that tinycoder basically built itself from there on out. i still use vscode to inspect the code sometimes, but i might replace vscode ultimately too.
If you think about it, facebook is just a ui over a database. Google is just a html form for a list of pages and hacker news can be replicated with a Microsoft Access.
If that seems stupid, is because it is. There are network effects and small UI benefits.
Dumb, fail for user freedom, nothing owned by OpenAI can be used to … create AI or anything that competes with them: scheduled AI, AI agents, AI tools, AI coding, chat, audio, image gen, video gen, shopping, and oh, anything the AI can do, soon social networking and hardware, what’s left that doesn’t compete with these assholes?
ChatGPT is a great breakthrough but it’s wasted if everyone has to worry about a noncompete with it. Seriously, how is it not insane to think we should outsource our thoughts and agree never to use the thoughts to compete with the thinker? Who wants to live in a world where nobody thinks and nobody can make anything competitive with their “Saviour Machine?”
Anybody who would join an org like that for a few billion dollars is a sell out. It’s an AI safety nightmare, too. I’m just flabbergasted millions of noobs accept not to compete with intelligence, wtf is this world, if you can’t use your thoughts to compete with your thinker, what is left for you? lol this is worse than black mirror
they have an infinite war chest and building windsurf/cursor isn't the hard part, building a brand and sales environment around it is. why risk failing the execution and losing focus when you can just buy one with momentum?
it's also a bit of multiple arbitrage in terms of what seriously addressing the developer market means for their valuation, they likely recoup the 3b instantly.
Sometimes it almost seems like the idea that LLMs are capable of instantly creating real, maintainable software is vastly overblown to inflate valuations...
Windsurf is terrible, I always use AI just in a normal website and I tried this product a few days ago and it asks me if it can run a command to make a file, which I find extremely strange, then it fails to write valid commands even to do mkdir.
That was my experience with OpenAI's Codex auto-coder thing (running o4-mini). It took 5 minutes and like 200 commands to do what Gemini 2.5 Flash (not even Pro!) did in about 30 seconds.
I see LLMs trying to do stuff that doesn't work in every AI coding thing I've tried, despite 20 pages of system prompts! (Or perhaps because of it.)
It worked allright for me when I was using it, a month ago. I cancelled because they somhow lost my paid credits and refused to refund me. No matter how great an AI tool, if the company is mismanaging user payments or usage tracking, it is useless.
IMO, there are few solid reasons to purchasing this tool
1. windsurf has lot of insights into how developer writes code, style, problem etc
2. for the prompt engineering that went into generating the code
3. only microsoft and cursor has the moat so they need to compete at the applications level not model level.
My prediction is anthropic, google or amazon will buy cursor. The next logical step to coding is building apps.
Maybe I'm missing something but I don't see many comments here talking about how this is a competitive move by OpenAI against Anthropic?
From what I've heard most people using/liking these agentic IDEs are using Claude models to power them, they seem to be the best at writing code. By buying Windsurf (and trying to buy Cursor) OpenAI can figure out why Claude is better at this task, fix GPT, and then make GPT the default for the coding use case.
Not sure it's worth $3B, but that's also not a lot to them when they can raise unlimited money at any time,
Value isn’t just the editor, it’s the workflow. Letting LLMs plan and act across multi-step flows is a hard problem, and Windsurf figured out a dev-focused version of that. Gains to be made in browser automation once you add structure, retries, and context. Feels like a bet on that pattern becoming default.
But yeah as others said, highly doubt that's $3B in hard cash, more likely a roll-up of shares etc.
Windsurf probably sees this as a win. I still think they're behind in some areas, Cursor's Agent feels faster and more responsive but Windsurf nails the rest. The documentation is far better, and the overall developer experience is more solid. Cursor still feels like a hacked-on plug-in in a broken VS Code fork. Even small touches, like built-in Linux install instructions, show Windsurf's polish.
Here are my two cents on cursors versus windsurf approach:
CURSOR shifted to a more agentic approach even for chat requests to reduce input tokens.
Previously, they used the good old RAG pattern with code dumps: Request with user added files -> Retrieval (when Codebase enabled) -> LLM requests with combined context from user and retrieval.
Now they seem to be doing something like this:
Request -> LLM with tools to search code base and/or user-added files
I get constant search tool calls even for user-added files. Big reduction in input token but I think performance suffers as well.
WINDSURF is still willing to dump code into the context, which gives them an edge in some cases (presumably at a cost of input tokens).
Windsurf is willing to spent to acquire customers (lower subscription cost, higher expenses for llm calls). Cursor has a huge customer base and is working on making it sustainable by a) reducing costs (see above) and b) increasing revenue (e.g. "Pro" requests for 0.05 with more input and output token).
Windsurf goes looking into the codebase and learning context before attacking the problem in my experience. Often Cursor tries it's best to just guess the solution without context and only really resorts to going deeper when you tell it they fails.
I find if I tell Windsurf to look at something it will, Cursor I sometimes lay everything out for it and it just doesn't even read it.
Ultimately though once you run out of requests on Windsurf it's very weak without Claude though, and the top up requests are burnt through too quickly.
I trialled Cursor for a month and then Windsurf. Cursor read entire code files in while Windsurf would read the first 100 lines (or was it 50?), then the next 100, and often stop before it got to the part of the file with the method in which was needed.
> Its the little things like having baked in instructions to install Windsurf on linux.
When I went to download Cursor the other day I noticed that they do not offer any .deb/.rpm packages and just offer the FlatPak (could be a Snap I'm not sure). This just tells me they really dont understand the community and just wanted to ship something for Linux and be done with it.
Windsurf has plugins for Jetbrains products, for vim, for emacs, for Visual Studio (not code), XCode, and even eclipse. They try to get as much of the market as possible, while Cursor focuses on the core functionality.
Whenever I tried Windsurf Editor, or their plugin for vim, and Intellij, it didn't feel polished at all. The basic function of autocomplete felt much much snappier on Cursor, and even on GitHub Copilot for vim/intellij.
>just wanted to ship something for Linux and be done with it.
This what all developers want for a platform. They can release their software and not have to worry about some "maintainer" switching out dependencies out from under them introducing bugs and crashes in what they shipped.
$3B for a fork of an IDE which Microsoft keeps crippling by the day by making it's best extensions not work with forks (eg. C++, Python, C#, Remote SSH, etc)..
That's a oversimplified view. It doesn't matter if it's a fork. It has customers and paying ones. And it has a brand. That's more than enough. $3 billion would be peanuts for OpenAI
VSCode must have over 100 times the user base of Windsurf and Cursor combined. All Microsoft needs to do is implement a halfway decent version of the context management features these forks added. That alone would be enough to halt user migration.
For users who've already switched to the forks, the cost of switching back is essentially zero, especially if Microsoft begins introducing changes that break fork compatibility. In that case, the migration direction would reverse almost overnight.
If it acquired those customers in an environment where Microsoft was not enforcing their marketplace terms it very much does matter if they have a plan for supporting plugins in the future.
Are Cursor and Windsurf going to ask plugin devs to push to their own plugin stores in addition to VS Code's? Will they rally jointly behind a single open store? They need to have an answer to Microsoft here, and for the good of the ecosystem I hope they do have an answer, but customers will flee quickly if they lose access to all the proprietary plugins and to the broader ecosystem.
> Are Cursor and Windsurf going to ask plugin devs to push to their own plugin stores in addition to VS Code's?
They should and probably will soon, and if I were them I'd even consider giving plugin devs a cut of their paying customer subs if MS gets competitive about it.
> but customers will flee quickly if they lose access to all the proprietary plugins and to the broader ecosystem.
Agentic AI coding is more important to customers than VSCode's extension ecosystem. VSCode is who has to worry in this equation unless they ship the same tools in the next few months and heavily subsidize them.
VS Code is shipping agentic coding in the form of updates to GitHub Copilot. I haven't used it extensively yet since they added agent mode, but it's obvious that they're gunning for this market hard, and if I were into VS Code I would not personally choose to lose the ecosystem for marginally better agent mode.
But that's a false dichotomy, Cursor is far from the only capable agentic option. Personally I switched back to using VS Code with Cline + Github Copilot (just for autocomplete and included model access to Gemini Pro 2.5/Claude 3.5/7 that I can use with Cline).
It's easy to downplay as a fork because it's such a young product but ultimately if people use Cursor or Windsurf instead of VSCode then it is VSCode that needs to worry about being upstream from them and Cursor or Windsurf making their extensions no longer work with VSCode.
does that mean that Windsurf will only support OpenAI models going forward? I doubt OpenAI will pay to have users use Gemini/Claude? Especially as all of these Ai coding tools (Windsurf, Augment, Cursor) are heavily subsidizing the users.
I wonder what Anthropic makes of this. Windsurf was like a top 3 customers of them, might be a big revenue blow too?
I think it's more of a time saver move by openAI - they can probably build something similar, cheaper – but, windsurf has established itself. Looking forward to see where this goes
they're paying to aquire places where you can sell tokens at a markup, because the future is multiple base models that are good enough for most user tasks where user gateways play the base model providers off each other and capture a lot of the value
probably a rare area I fully agree with HN on– the IP here seems weak and it's not hard to swap out code editors, nothing like tearing out Salesforce or other sales-driven tooling. and idk if first mover advantage actually means much in the next 10 years given how dynamic the underlying models are.
but undeniably these cos are all a great lesson in just how much cash lies in executing first/near first
Maybe time? OpenAI has access to basically infinite capital right now, if they believe this will be an importnat market and they could save a few months on launching this acquisition may be worth it for them.
I've found Windsurf more reliable/efficient than any other editors by leagues. How ever they have named the tools, crafted their prompts and generally how their internals reason is just on the money. I don't think that is easy to replicate, iterating on prompts over product releases whilst not pissing off your user base constantly is a feat in of itself.
People seem to be pretty negative about this but of all the AI dev tools I've evaluated it's the only one that's felt meaningfully better than just using the web interfaces of the various frontier models.
I don't think it's good value for the money but pretending it's just a VSCode fork that wraps LLMs is underselling it. There's something they're doing that makes them better than Cursor, Claude Code, etc.
Would love to know what exactly Windsurf brings to the table that justifies $3B. Infra play? Specialized team? Or is this another move to consolidate talent before others can?
Pretty stupid move with Microsoft moving to put the kibosh on all of these proprietary vscode forks. Could be worth almost nothing in a matter of months...
With $3bn budget you can replicate it in few months, promote for free using your own stronger brand and you're left with roughly $3bn in the bank to do whatever you want.
Your sentiment is very common, which reminds me of those days where everyone claimed to be able to clone twitter/airbnb/dropbox/gmail/whatever over the weekend.
The real value lies in a successful execution, Windsurf is live and already has brand popularity with revenue stream. When it gets acquired by a titan, the die hard fans(vibe coders) will gain new trust that the product is not going anywhere and instead has solid future(we don’t yet know if oAI plans to do Msft style acquire, extend, kill yet). Cumulatively, this actually increases the value. Also Windsurf already has established enterprise revenue, hence brand trust and experience is already there.
In summary, an existing live and proven product is worth more any day over 100% uncertainty of building a sufficiently capable team to perfectly execute the same idea in specific deadline and also have the added burden of marketing, market penetration, user acquisition etc.
Github copilot, vscode and apis through azure - basically everything through Microsoft - is.
Alternative to Microsoft's monopoly in enterprise that exists is open source.
Comparing this situation to twitter is more like if there was some chat api service, known more than twitter itself, that twitter is using/wrapping where other alternative clients exist, some with stronger popularity, some being open source.
here is the thing, even those editors are relict of the pasts, the code is still in the center in these editors. thats something we need now, but not in the near (2 years, 5 years, 10 years?) future.
then the prompt is the coding, the reasoning is the execution, the code just an abstract layer that we do not care to much about i.e.: like assembly, machine instructions.
we know it exists, bit even here on hackernews i would guess only a small fraction know how it really works on a detailed level.
there will still be coding, instructions (prompt) -> execution (reasoning and AI code and code execution -> feedback (debugging to AI then and one point to the user)
bur actual looking at the code, well, thats only when this cycle annoyingly fails.
so current IDEs are still built from an code first mindset. this will not be the IDE of the future.
That makes a lot of sense, for the customers, but OpenAI is not profitable on even their $200 subscription. I doubt Windsurf is turning a profit either.
Buying a "bundle" should result in a lower price, as compared to buying both tools separately, making the loses worse. Unless they can reuse some of the same infrastructure and save a lot of money that way.
They seem almost exactly the same as Cursor, but even using the exact same rules, Cursor gives much better results than Windsurf (which performs below viable for me) - my test case was a complex Python project.
I like Windsurf for RSE, but it sometimes gets a little too excited which can take me out of the flow to undo stuff and get back into the groove of things.
The Claude integration is quite nice - I hope that doesn't take a step backward with the acquisition.
I don't get why people want the AI right in their editor. In another windows inside the editor, fine, but not inline with code I'm writing. It's super distracting to have AI auto complete pop up at random all the time. As always, typing speed, or speed at generating raw code, is not the bottleneck in programming. The crux remains design, in which case having the LLM on the side is just fine (if you use it for that).
There are some niceties about inline completion (like spelling out a log message that's obvious from the surrounding code) but I don't get the hype much beyond that.
If you have tried the completions in copilot, you are right. They are complete garbage.
Windsurfs on the other hand are much better. The only issue is that windsurf is super aggressive about them, but it is able to do do things like "the user made a change on this line, he most likely also want to make the change here".
AI autocomplete is the best thing I've experienced in developer experience in my career since git won over subversion.
I don't use LLM code prompting, but autocomplete is my jam. It's getting things right 90% of the time when I'm plumbing fields or refactoring. It makes life so much more pleasurable, and I say that as someone who is already using a statically typed language with robust IDE refactoring capabilities.
Good for them, always rooting for startups who win.
That said, I have tried Windsurf multiple times, and it wasn't a pleasant experience compared to Cursor, which I've been using for more than 6 months as a paid customer.
However crazy the 3 billion valuation is, windsurf's valuation is still very sane compared to that of Safe Super intelligence, who exists for less than a year, with no product, no roadmap,and virtually no hype, but is worth at 30 billion.
I use ChatGPT’s “work with” code helper and one of my biggest uses on ChatGPT. It’s a good first line before I pull out the big guns(APIs). Sadly the code canvas is rarely as it’s geared mostly for single page web app functionality useless demo tests. Maybe this is where Windsurf can come in
I cannot pretend that I know what is going on - I don't.
I think the long-term play here is something to do with Agents and they are simply cornering the market because coding tools are part to it.
That being said, quick search around what people are building with these VIDEs reveals mostly landing pages that are actually not even that good. For the amount of money spent one could have easily bought a good template or pay someone to customise an existing one.
Very surprising outcome, since OpenAI went after Cursor (twice) [0] And I originally thought that Cursor would be bought instead a day before the rumour [1].
It was smart for Windsurf to take the offer and to get greedy in this hype cycle. Unless Cursor is thinking that Anthropic or someone else will buy them for a lot more, its going to get extremely competitive as the switching cost for Cursor is zero and that ARR can disappear very quickly.
Copilot will attempt to destroy Cursor on price and functionality for however long they want to.
Very risky for Cursor at $9B valuation (which I think is overvalued and based on VC FOMO).
Probably the other way around. Windsurf and co (Aider, Cursor) drives a heap of traffic to their API from which OpenAI actually profit. They just need to have their own tool to lock customers in their ecosystem.
I need someone to convince me this isn't one of the biggest waste of money on an acquisition. If OpenAI can't build an official IDE for less than 3 Billion then what are they even doing? Windsurf can't have that high of a userbase that you feel the need to pay for it.
These deals are mostly in stock, not cash. $3b cash is not something most companies can afford to part with, and additionally, making deals that are stock-heavy creates an incentive for the leadership of the acquired company to keep working towards the general interest of OAI, and not instantly retire.
I'd define them as a large company acquiring a private startup. Slack >> Salesforce was a cash-and-stock deal. Postmaters >> Uber, all stock.
In cases where the company being acquired is already publicly traded, those deals would have to be all cash as their shareholders would need to be bought out. IBM paid cash for Hashicorp, and Doordash will acquire Deliveroo in cash.
> I'd define them as a large company acquiring a private startup. Slack >> Salesforce was a cash-and-stock deal. Postmaters >> Uber, all stock
Okay, in that category of M&A in practically any category, the vast majority of deals are all cash. Deferred, for executives, in most cases. But cash.
> In cases where the company being acquired is already publicly traded, those deals would have to be all cash as their shareholders would need to be bought out
Not true. Preferable. Easier. Not not a requirement.
instinctively I agree but it's all about timing: if they try to build their own IDE and hired people it would probably still take a couple of years to get a decent product. I don't know about patents.
> if they try to build their own IDE and hired people
Oh, haven't you heard? Hiring people to write software is so last decade. Maybe they just didn't want to vibe code a Windsurf implementation and decided to buy a press cycle for $3B
This is classic OpenAI - acquiring competitors rather than innovating internally. They're desperately trying to keep up with competition from Anthropic and Microsoft's GitHub, but throwing money at the problem is hardly a creative solution.
What's especially rich is the timing - right after OpenAI backpedaled on their restructuring plans due to "public pushback" (read: Sam Altman making yet another governance blunder). Now they're dumping billions into a tool that's essentially the same thing everyone else is building.
Who are these people that give OpenAI all this money? Aren't Microsoft, SoftBank, Nvidia publicly traded? Don't they owe a fiduciary duty to their investors? I'm surprised it's legal to just hand over a blank check to random private companies to make nonsense purchases. This isn't going to end well.
If I were any of these companies I would be suing OpenAI to try to get my money back. Thrive, ARK, Tiger and the others can pound sand
https://archive.md/l6n9H
Incredible timeline to a $3B exit
> Windsurf began in 2021 as Exafunction, founded by MIT graduates Varun Mohan and Douglas Chen. The company initially focused on GPU optimization before pivoting to AI-assisted coding tools, launching Codeium, which later evolved into Windsurf.
> Series B (January 2024): $65 million at a $500 million valuation.
> Series C (September 2024): $150 million, led by General Catalyst, at a $1.3 billion valuation.
> May 2025: $3 billion acquisition from OpenAI
I wonder how much of the value is really from the model or the tooling around it. They all use the same models (mostly Claude, others have been horrible and buggy in my experience). Even co-pilot agent mode now uses Claude. The editor has their own LLM (?) that does the apply since LLMs often return snippets. They work well enough on Cursor. And then you have the auto-complete, which I think is their own model as well.
But the main value from me is from the agent mode and 95% of the value is the underlying model. The other stuff could be more or less a VS Code plugin. The other benefit is the fixed pricing. I have no idea how much 500 calls cost if I were to use the API, but I expect they're probably losing money.
Incredible timeline - also helpful to understand the OpenAI side.
1) OpenAI is valued at 300B (as of March 31st) https://openai.com/index/march-funding-updates/
2) OpenAI recently raised 40B from SoftBank and others.
3) Windsurf is getting roughly 1% of OpenAI's valuation.
OpenAI needs to keep moving fast to outpace MS, Google, and others -- and I think we can all agree that agentic coding is a major trend -- that is likely to keep growing really fast -- and super high leverage in that the folks doing the coding are well paid -- and more likely to be early adopters than any other field. (e.g. if openAI wants a fast way to grow beyond $20-$200/month, owning a tool like windsurf is a good move)
Some folks have been speculating the cash/equity split. I'd be confident whatever number they arrived at de-risks things for windsurf, and preserves the right amount of cash on hand for openAI.
Even if OpenAI is burning 10-20B a year, with the recent raise would buy them between 1-2 years, and given the pace of AI development that's a pretty long time.
I’ve been a WindSurf customer since day one. It was my first true AI agentic experience.
[Dev mode] While working on Alembic migrations I broke one of my migration files. After an hour of manual debugging I handed the task to WindSurf. It methodically checked every config file, applied the migrations one by one, and narrowed the issue to a single file. It rewrote the migration, verified the fix, wrote tests, ensured everything passed, and opened a PR. I reviewed it and it worked flawlessly.
Regarding the acquisition I don’t understand why OAI would pay $3 B. The team is strong, they have lots of data, and the agentic flow is great, but all of that feels commoditized.
Claude Code launched two months ago and I prefer it to WindSurf, Cursor, and Aider. Augment Code also ranks above WindSurf for me.
If I were in Sam’s place I would have doubled or tripled down on Codex CLI. Just my 2 cents.
I was involved in an M&A once; my role was to evaluate the technology and determine how long it would take us to build a competitive product. If it was less than some X then we’d build it, greater than X and we’d buy. The function for X was not clear to me from my perspective; it had legal fees involved, etc.
The person leading M&A said an intangible aspect of the price is what it does to the adjacent market. If the leading product A is valued during a raise at $Y, and you buy the next best product B at 1/10 that, you cause future issues with raises for A.
Might this be an attempt to clip Cursors wings?
That's a really interesting thought, I'd love to get involved in software PE/M&A on the technical analysis side but I don't have the academic pedigree for it (it seems every shop that does this work is 90% Ivy and Ivy-adjacent universities and FAANG-level work history).
So if I'm understanding your point then part of the value in paying $3B for Windsurf is that it acts as a pricing anchor on future raises (and presumably acquisitions as well) for competing products? So Cursor is less likely to raise at a $30B valuation if Windsurf is 95% as good and just sold for 1/10 that.
The insanity of it all is that these companies are worth about $3M.
I would also think that a critical component of X there would be the opportunity cost of time spent on building in-house while competition chugs along.
> WindSurf, Cursor, and Aider. Augment Code also ranks above WindSurf for me.
Bring on (a lot) more competition! I am waiting for the point where "Simple Pricing" (Augment Code has that on the pricing page) means fixed pricing; Simple is NOT '600 messages included' because it's impossible to know what the ROI from those 600 messages is, so it's very far from 'Simple' (many of those prompts will deliver nothing or, worse, having to rollback because the agent produced garbage). I know it's not sustainable, but the only thing that will keep me not jumping from one to the other, signing up with different emails, trials, coupons etc is if they will lose the restrictions on usage. They will, as they have to compete and it's worth it seeing this acquisition; losing 10s of millions a month to get/keep people and getting nice growth is what they do to get the billions. So bring it on!
If he's (Sam) trading equity on a grossly inflated OpenAI for the acquisition then he's likely not paying real money for the company and thus he is expanding his roadmap for cheap.
sometimes companies are acquired for things the public has not yet seen.
Companies are acquired for customer base, ARPU, and growth. Same criteria as when when raising funding.
Those are the best reasons, but companies are also acquired for marketing, hype, to relieve a sense of fear, to curry favour, etc…
I don't think that's the case here. Windsurf wasn't leading the agentic coding market. They were doing a decent job but others are bigger. Cursor has the brand recognition and Claude is getting a lot of recognition too. MS has github copilot which is still a good brand and Google has been catching up with Gemini.
OpenAI has a new thing called codex but it isn't very good yet. I tried it and it's super flaky. Lot's of errors and it gets stuck when that happens. OpenAI needs something good urgently because agentic coding is the key AI feature right now and the blue print for non coding agentic solutions later. Cursor is probably too expensive currently and windsurf looks like their models are a bit better.
So, OpenAI gains something they don't have: a credible developer option with an active user base and some core IP in the form of training data and know how as well as custom models that they can fold into openai.
3 billion is a lot but not if you consider that world + dog in the enterprise world will be spending big time on AI subscriptions for their developers. This stops being optional in 2025. Millions of developers will be on paid subscriptions permanently very soon. If you start a new job you can expect to get a laptop and a paid subscription to whatever is the agentic coding tool of choice in your new company.
OpenAI wants double digit percentages of that revenue. 1M users paying something like 50$/month would amount to 600M revenue per year. I think the prices will go up and the amount of active users as well. Reason: as these tools are getting better they start saving non trivial amounts of engineering time. At that point you have to value the tool in terms of developer cost. Not 1 to 1. But it's worth a sizable portion of that.
I work in a small startup as the CTO. This is an no-brainer for us. We're cash strapped so we only spend on important things. This would be one of those things. We're doing things I previously would have needed to expand the team for because I would have had no capacity to do those things in the current team. So, in terms of value for money spending on these tools is easy to justify.
I get lots of people are skeptical about AI stuff here. But I would say that a lot of those people suffer from a short term focus and bias. Three years ago none of this stuff existed. Now it's a multi billion$ market that is set to grow rapidly. Stuff is getting better at a very rapid pace. Just stating facts here. 3 billion is a bargain if Openai can make this acquisition work for them. They are buying time to market here. They don't have a year to figure it out. In a year or so this market will be carved up and locked into hard to change year long SAAS contracts. At that point getting people to switch tools will get harder and harder.
> OpenAI has a new thing called codex but it isn't very good yet. I tried it and it's super flaky. Lot's of errors and it gets stuck when that happens.
I agree with this, not sure the experience of everyone else but I felt like Claude Code is more useful.
Meanwhile, I'm keeping tabs on Aider and open-codex, what other options are there?
Thanks for mentioning open-codex. Did not notice that there is a codex fork which is open to other models (update: totally missed that original codex allows that too now). How do you like it? Especially in comparison to Claude Code?
I wanna pick your brain a lil. Are you saying agentic AI has helped you replace devs that you would otherwise need for your startup?
In your opinion, what should one do/learn to get a SDE/related gig now? What do you/other companies look for?
Not replace but it allows me to scale what we do for things we previously would have dropped because it would require growing the team, which we can't really afford. It's a case of getting a bit more out of developers in terms of quantity and scope (mostly this) of what they do. Not a full developer but enough for it to be meaningful. But it's not nothing either. Worth paying for. AI is a lot cheaper than a developer is so I don't need to replace my developers. I prefer people that are multi disciplinary and able to pick up new skills as they are needed. Agentic AIs are good for that because they give you enough to work with that you can get productive with whatever you need to wrap your head around in little to no time.
Companies can be a bit slow to update their hiring processes to their needs. But good developers should be ahead of the curve in any case. For this, just be proficient with the tools.
Be ready for the inevitable interview question "so, AI ... explain me how you are using it and what you are doing with it?". Much easier to answer that question if you have some meaningful time of routinely using this stuff behind you and can articulate what works and doesn't work for you.
And if they don't ask, that's actually a great question to ask back if you get the opportunity "I've been using agentic tooling, how are you guys using that a <company name>? Also I would like a subscription to <my favorite AI tool> if I work for you". Stuff like that makes you stand out as ambitious and interested in the future. There are of course going to be places that maybe don't like that. But then ask yourself whether you'd want to work there. So, either way, you learn something.
Thank you for responding :)
So, would you consider it a bad idea to get into web dev, more specifically backend and infra?
Do you think using LLMs can accelerate learning software dev and programming skills?
I would look forward to the next 20 years and not backward to the last 20.
The whole frontend/backend distinction did not really exist until the web. And infrastructure is definitely something that should be automated far more than it currently is. If it needs babysitting by a team of devops, you just created a lot of work rather than automating/solving it. Tedious and repetitive. It has "AI will make this a lot easier" written all over it.
So, just be ready for the ambition level to be raised for developers. Learn to build the whole system, not just bits and pieces of the system. Lean on AI to get stuff done and figure things out. It's all just code. None of it is really that hard. But it can be a lot of work if you do all of it manually.
And let's be honest, agentic tools are showing promise and great progress but they are nowhere close to independently working on existing code bases. That's not how I use them. But they are great for problem solving, debugging, prototyping, exploring some new languages and APIs, and generally taking care of more tedious coding tasks.
In this case, maybe it's an acqui-hire?
doubtful imo
Because they have users and OpenAI has seen the massive drop off in coding usage since Claude Code came out. My personal Chatgpt decline is at least 99%. It’s also 1% of their current market value. So not really a big deal.
They're paying $3 billion because money is hyper plentiful for OpenAI at present. Basically because they can. Money isn't their problem right now, it's not a scarce resource (maybe it will be in the future of course). They're trying to capture and lock-in, so as the hurdles and regulations go up they're one of the huge winners left standing.
Try replacing Uber today, it's impossible. Nobody is going to give you billions of dollars to try to do it. It'd be an absolute nightmare to attempt it.
Uber has already been replaced, at least in some parts of the world. We recently went on holiday to Malta and on check-in the hotel staff told us not to bother with Uber, Bolt worked way better and had more drivers (Bolt is a European Uber competitor based in Estonia).
So we signed up for Bolt and sure enough drivers were plentiful, the app worked great and there was no downside over Uber. I'll certainly be trying it again in future in other markets.
The reason Uber invested in self-driving cars for years is that otherwise they have no sustainable edge. It's just a taxi company, which is a low margin business. People who can make slick mobile apps are plentiful and it takes a minute or less to sign up for a new service. Uber grew to its current size by buying market share using investors dollars, which was always a time-limited strategy. Once they started having to turn a profit prices rose and their edge over their competitors was lost.
Uber feels like such an apt comparison to OpenAI to me. The service they provide is obviously going to be absolutely huge, but no guarantees at all that they’ll win it or be last man standing. I don’t see a world in which generative AI doesn’t continue to be a massive disrupting force, but no particular reason to think Anthropic or OpenAI will still be independent entities in a few years.
I’m even more bearish on Uber than I used to be, as someone who’s used Grab and Careem and Bolt extensively, and seen Uber have to beat a retreat from SE Asia. If their more nimble competition get a foothold in the US they’re toast.
I feel like the value-proposition of Uber was three-fold.
1. Solving a pain-point of many people re: hailing a cab, via an app that works everywhere.
2. Using VC funds to (initially) pay drivers more than you, the customer, were paying them.
3. Ignoring local regulations and passing the savings/convenience on to you.
1 is nice but I don't think they established much of a moat (both drivers and customers are willing to use multiple apps); 2 isn't sustainable in the long-term, and they failed to leverage 3 to establishing a permanent right to operate as they had been in most markets.
I think this makes Uber an even more interesting benchmark for other unicorns, since besides "solving a real problem without establishing a moat" they are also often burning through VC cash to prop up their business model while ignoring some laws which they may not be able to get away with ignoring long-term.
1 especially is a social function. Having to have a million different apps is terrible, but if there is too much competition for drivers it's inevitable to churn through apps because of marketplace pricing and rent seeking on all sides
Is there a writeup or a recording. Would be nice to follow through
talented and smart folks for sure but can't not notice how much luck it is especially because its like 100% just better models. Windsurf raised a ton of money and then said they pivoted which they had millions raised to just do something completely different that likely wouldn't have been easier to raise for. Even in an interview with the cursor founder he kind of dumbly rambles that they launched and then basically lost a ton of traction until GPT4 came out. They have some core features like autocomplete but I'm struggling to see vision other than getting training data for iterative dev is a partial moat compared to just seeing commits and final code bases.
Training data is almost certainly their main reason for this acquisition. Users themselves and the models they use don't really matter. What matters is their interactions with the models. Especially if you're trying to build coding agents that will be marketed to companies for $10k a month. OpenAI is going for the industry B2B opportunity here, not consumers or end users.
But aren't they getting this data already at a much larger scale? GPT is still one of the backbones in many coding assistants, even Windsurf.
It's a bubble about to pop. That's where the value is coming from.
5 years ago if you said coding tools would be worth in the billions in value it would of been surprising to most people. Dev Tools were the thing you could never get a company to buy for you or were just free for most people. Interesting times.
Dev tools are still very hard to sell (I know, I have a dev tools company). Claude Code, Aider and Codex are given away for free. What people are buying is access to proprietary general purpose models.
Bubble or not, given the exit, Windsurf's (Codeium) focus on enterprise sales motion has been rewarded rather handsomely: https://research.contrary.com/company/windsurf / mirror: https://archive.vn/ThWNz
Yeah, in the recent Lightcone Podcast episode, Varun was talking about how they have a lean eng team but large sales org. I thought that was super interesting for a dev tool since I was expecting a dev tool to involve bottom-up sales to the dev instead of top-down sales to a leader like a CTO or VP of Eng
It will be super interesting to see how they do against the inverse: an engineering-focused company that wants to win devs from the bottom up
100%. You're referring to Cursor?
It's all about stocks
Totally agree - a lot of the "magic" still feels like it boils down to whoever has the best underlying model
I think part of the value is customer acquisition rather than product.
The value is in the prompts being sent to OpenAI. Massive training depository.
Only thing better would be a social network, which supposedly they're working on.
But OpenAI already knows every single prompt sent to its models. They don't need to buy Windsurf for that.
That's a bit like saying having access to Google is as good as being Google.
All they really see as a model provider is little fragments of the picture, like trying to reconstruct the Mona Lisa by knowing which paint swatches Leonardo used.
In other words, they only saw whatever Windsurf sent as context with a "fix the bugs" prompt stapled to it.
By owning Windsurf, they see the entire source code of what's being built, all the time, plus how the model is interacting with it.
There's a massive amount of value in what happens client-side, and behind the scenes. The "director's cut" of context.
Huge difference.
So just put up together a comparable VS Code based AI IDE in a couple of months and bundle it together with the ChatGPT subscription? They'd get loads of users very fast..
I think it is exactly this. There is no doubt whatsoever OpenAI can do this, but they decided not to. The reason, I think, is that they don't want to be a couple of months late. In other words, they spent $3B to save a couple of months.
There's much more to be gained if you also have the client side of those interaction. You can get signals from "accepted" completions/plans/etc, number of edits made to those completions, how users use context, what was passed in context from a code base, and so on.
And that's just on their models. They'd also get (at the very least) signals on their direct competition, if not straight up prompts+completions as well.
Now they also get to see what is sent to Anthropic, Google, DeepSeek, etc., and what is returned. At scale, for a prime area of concern.
That's valuable user and competitor data actually.
Reread the comment I replied to. It has nothing to do with Anthropic or Google or DeepSeek.
> Incredible timeline to a $3B exit
dot-com vibes. Maybe not quite the same as Pets.com but still...
The right time and the right place, plus they did the work, ofc; but I'm sure 80% of this site has worked as hard as, or even more, than what it takes to clone VSCode.
I'm jelly. Very rarely you see in history someone lucky enough to be riding the absolute top of the wave. Even OpenAI took about decade to cook their breakthrough product.
Visual Studio Code Agent Mode uses whatever model you tell it to use.
It's a beautiful world where you'll only put a little over 220 MM in and get 3000MM out mere months later.
The value is the team and their thinking and the customer base.
How is it different to a bunch of other apps which and tools which offer more or less exactly the same?
> I wonder how much of the value is really from the model
> The other stuff could be more or less a VS Code plugin
The other stuff would take a team 6 months to implement. This is where the valuation comes from. Time to market, they are there TODAY.
6 months of anyone's time is not worth 3 billion dollars.
That was my thoughts too. No text editor is worth $3B, and probably not even VSCode is. So I think this deal was about buying more customers/users and buying "relevance". OpenAI lost it's monopoly and they're worried they might become irrelevant so they basically just purchased something popular to remain relevant.
it's not just one person though
As usual, HN misses the point. The customer list was probably worth about $2.99 billion, and the engineering work about $10 million.
> customer list
How many customers do they have? At $30 per month it would take forever to pay off even with a lot of growth.
Open AI could release an equivalent VS Code clone and make it entirely free and it would still be a lot cheaper than $3 billion.
Is OpenAI having trouble acquiring enterprise customers?
What makes you think $3B worth of customers were committed to Windsurf at all, much less in a sticky/exclusive way?
I'm skeptical about this VSCode fork commanding a $3 billion valuation when it depends on API services it doesn't own. What's their moat here?
For comparison, JetBrains generates over $400 million in annual revenue and is valued around $7 billion. They've built proprietary technology and deep expertise in that market over decades.
If AI (terminology aside) replaces many professional software engineers and programmers like some of its fierce advocates say it would, wouldn't their potential customer base shrink?
Professionals typically drive enterprise revenue, while hobbyists—who might become the primary users—generally don't support the same business model or spending levels.
What am I missing here?
Part of what you're missing is that OpenAI needs to justify its own overinflated valuation. They raise money on the premise that an AI-native company can and will outcompete giant established players, so lowballing Windsurf would run counter to the narrative they're selling to their own investors.
The article also doesn't say that it's $3B in cash that OpenAI is spending. They might be giving Windsurf $3B worth of OpenAI shares - paying an inflated value for Windsurf with their own inflated value.
OpenAI just had a fundraising round that put them at $300B. Maybe they're just giving Windsurf 1% of OpenAI. Maybe they're even giving less than 1% - if OpenAI was worth $300B at the end of March and $150B last October, maybe they're worth $400B now. Maybe Windsurf is getting 0.75% of OpenAI that's "valued" at $3B.
> OpenAI just had a fundraising round that put them at $300B. Maybe they're just giving Windsurf 1% of OpenAI
That is the most hilarious maths I have ever seen, if this is true then it's maybe the biggest "holy fuck it's a bubble all the way down" I have ever seen
So they are effectively blowing their own bubble?
That is what it looks like from where I sit, yes.
They built all of this assuming VSCode was a solid foundation for the next 30 years and I've completely undermined VSCode's technical foundations. Their castle is gonna sink into the swamp...
JetBrains makes $400M in revenue and is 10+ years old. Cursor is 1 year old and makes $300M in revenue.
One is going to be valued at a much higher multiple than the other.
$400M in real revenue versus $300M in annual recurring revenue (ARR) are totally different things. Real revenue is money actually earned, while ARR just multiplies one month's subscription revenue ($25M) by 12, ignoring customer churn.
Startups love flashing ARR figures because "$300M ARR" sounds impressive, but without knowing churn rates, they might never actually collect that full amount.
JetBrains however collected real $400M in a year.
I’m spending more on Cursor every month. Worth every penny. I’ve never given a dime to Jetbrains.
Unless you have reason to believe the revenue is declining in recent months or will decline in near future, ARR is a better metric. last year real revenue made sense only for low growth companies.
Cursor just lost access to the extension marketplace and key proprietary plugins that they were using against Microsoft's terms, Windsurf has been eating a chunk of their mindshare, and Copilot is catching up.
That's three good reasons to believe that lots of people will be cancelling in the next months unless something changes.
The entire reason OpenAI has a high valuation is the expectation that AI will get a lot better in the next few years. If that happens, building a clone of Cursor/windsurf should be trivial. The only reason you would buy windsurf today is to either pump up the bubble OR use it to increase your market share of developers by taking users away from claude
I feel jetbrains is squandering an opportunity here. Cursor is significantly easier to build then any IDE in the jetbrains ecosystem. The technology jetbrains is very hard to replicate. While the technology cursor uses should be trivial to replicate.
If jetbrains can combine there IDE technology with cursor technology, that would be ideal.
I think the problem is jetbrains tech is sort of already very biased in a certain direction and it's hard for them to pivot as fast into this new AI direction.
JetBrains launched their cursor competitor a few weeks ago.
I prefer Claude Code still because it has access to more tools - Junie seems unable to fetch URLs and do other things. But that's a tiny gap that JetBrains can close quickly, and the Junie UI is quite pretty. Plus, inside the IDE they can equip the model with far more advanced tools than Claude Code will have from the CLI: inside Code Claude has to explore the codebase by banging stones together with ripgrep, whereas in the IDE it can be equipped with tools to access the indexes and navigate around like a human would.
In theory, JetBrains should be able to compete very strongly in this market. Their single line completion model is already excellent.
I trust Jetbrains. Only company I really trust. They work for me, and have showed it again and again.
They literally have that it is called Junie and after comparing cursor to it we settled for Junie as it does a good job with rust unlike cursor.
> If jetbrains can combine there IDE technology with cursor technology, that would be ideal.
Just give them some time, they're not stupid. I'd drop Cursor in an instant once JetBrains catches up, because IntelliJ IDEs are just a way more powerful.
Didnt Jetbrains launch their AI last year, is anyone using it?
They're giving out 1 month free if you're paying for their IDEs already. I've tried it last year and it was very limited, not "agentic". Now they've launched an agentic version called Junie and also gave another 1 month free, and I've tried it again.
It's a nice improvement over the last edition, but still quite not "smart" as Cursor or Windsurf. The agent seems too shortsighted compared to competitors: it may stop looking for files or making edits sooner and you're left with code made with incomplete context (that does not work or just doesn't address your needs). It also does not fix linter/compiler errors from its own output code before finishing, unlike Cursor.
Try out Junie
"I'm skeptical about this VSCode fork commanding a $3 billion valuation".. Nothing to be skeptical about. The market has spoken. It was worth 3b to OpenAI. Companies arent worth a vague notion of what "value" someone in an armchair thinks they might be worth, they are worth what people are willing to pay, and OpenAI paid.
> If AI (terminology aside) replaces many professional software engineers and programmers like some of its fierce advocates say it would, wouldn't their potential customer base shrink?
This is such a good point. The best reply available to the AI hype-men would probably be that LLMs "democratize" coding and therefore that even more people will use IDEs in the future, but that sounds like BS to me -- not unlike AI/hype itself.
indeed, and that is why you see adobe declining, because, their customer base is shrinking even as they add AI into their tools.
Just consider what it fundamentally is: a company at the leading edge of a product category that has found absurdly strong technology/use-case fit, and is growing insanely fast.
Looking for a moat in the technology is always a bit of a trap - it’s in the traction, the brand awareness, the user data etc.
> Looking for a moat in the technology is always a bit of a trap - it’s in the traction, the brand awareness, the user data etc.
Traction, brand awareness, and user data do not favor Windsurf over GitHub Copilot. The few of us who follow all the new developments are aware that Windsurf has been roughly leading the pack in terms of capabilities, but do not underestimate the power of being bundled into both VS Code and GitHub by default. Everyone else is an upstart by comparison and needs some form of edge to make up for it, and without a moat it will be very hard for them to maintain their edge long enough to beat GitHub's dominance.
Definitely take that point. But this valuation is perhaps more about how much that traction, brand and data is worth to OpenAI, who cannot buy Copilot. $3bn doesn’t seem so disproportionate in that context especially given the amount of money being attracted to the space.
But copilot is bundled and is free, and it's still losing to cursor
Define losing? My company pays for Copilot but not for Cursor, and it's not at all clear to me that we're the exception rather than the norm. What numbers and data are you working with?
Copilot has every incumbent advantage, so if Cursor is doing halfway decent in the market (which it is), it's winning by default
Incumbent advantage of being in VS Code already? Thing is, Cursor is basically just VS Code, there's hardly any barrier to switching, so it's quite a weak advantage.
That's not actually how unseating an incumbent works. The incumbent can adapt to the threat for quite a while if they act on it, they just have to not be Blockbuster. Copilot is showing every sign of making up ground feature-wise, which is bad news for the runners up.
In brand velocity maybe, but copilot is rapidly reaching feature parity with cursor and will invariably overtake it—while costing less to users.
Same with Google vs OpenAI. I tend to agree with the sentiment that I most frequently hear which is that OpenAI is the currently popular brand, but that can only carry them so far against what will eventually be a better offering for cheaper.
High valuations for companies you've never heard of with no moat - it comes down to cronyism/nepotism/fraud.
Yeah it seems like there's really no "adult supervision" at all in OpenAI. This purchase was a panic move. Windsurf would be worthless without the AI. Probably OpenAI knows that AI is now a commodity technology and no longer a space they can monopolize so they're just trying to get off a ship that's sinking, and find some viable path to having a tech that doesn't ultimately depend on OpenAI even having a monopoly any longer.
They have a healthy enterprise customer base, and an engineering team that clearly knows how to work with power users (which OpenAI is bad at).
OpenAI buying a company that is dependent on their competitor.
Cursor purports 200m in projected yearly revenue. With some months having 40% month over month growth. The trajectory is vastly different.
Whether or not it's justified is a different matter, but for startups valuations are more about potential then current performance.
Cursor purports $300M in annual recurring revenue (ARR) but stays silent on churn.
They made $25M from subscriptions one month, took that number, multiplied it by 12, arrived at $300M and everyone has been running with that line without ever asking what their churn looks like.
They could have churned $24M the next month, ask yourself why they are silent on churn if they are doing so well.
Venture capitalists aren't ignorant, their business revolves around knowing exactly what churn is. Cursor has raised $1 billion with a $9 billion valuation. VC's willing to put in that much money has looked at their data and knows what the retention rate is.
If their plan is to make their money back selling the company, then they don't care about revenue or retention rate. The company just need to look like it might be doing well.
No, venture capitalists aren't ignorant, but their goal also might not be to build and run a healthy company long term. It might be to turn a quick profit by selling a startup to another company.
They have no moat, Cursor does the same stuff. Microsoft's moving to kill all of these anyway and has added agent mode to copilot.
OpenAI would have gotten more value by setting that 3 billion on fire, at least it would have powered the data center for a little while.
OpenAI needs a product team
hiring is hard
it's a high-functioning team swimming in contemporary design and eng practices
code is emerging as an important battleground
OpenAI has the $$$
I bet they can hire best minds in the world for a fraction of $3 billion.
If that's so, then why is Codex such an inferior product to Claude Code? And why haven't they already built an code editor or at least VS Code extension yet?
JetBrains has been making IDE for a decade. They were the only company that actually made money by selling IDE. So I assume they have the best programmers who understand IDE.
However they fail to make a Cursor competitor so far. This alone suggests it's a harder task than meets the eye.
It is, but you are assuming that only a well known IDE team would do it. To me JetBrains is the least likely to be an innovator here because they depend on their reputation for being a mature technology.
Someone like me isn't known at all but it means I have been able to experiment for a long time without pressure, which is how you do real innovation.
JetBrains as a company probably owns 10 million lines of code and it's just really hard to move fast when you're tugging that kind of ball and chain
It is ironic that the company said to be cooking AGI is acquihiring software engineers because they can't develop it in-house.
> What am I missing here?
That AI is in a bubble akin to the crypto craze from a few years ago, and the valuation of these companies is divorced from their underlying business fundamentals
> when it depends on API services it doesn't own
It now owns the API services.
Well you see Jetbrains is a European company unlike the super special boys running an inherently more valuable American company.
Steel man: Windsurf own the customer relationship. The models are just generic interchangeable services they use for processing.
Realistically: I don’t know how many users windsurf actually has and I never actually met anyone that uses them. Whereas Cursor AI took a huge percentage of the VS code users I know in real life.
I use Windsurf. It had the smoothest agentic experience when I subscribed. I think still does.
If OpenAI just provides AI, then the various IDEs development wrappers / IDEs / low-code etc. can collectively bargain against OpenAI for low rates. If OpenAI has an alternative, then they can charge higher rates for all plugins/ etc. and give the market an alternative.
If enterprises require fewer software engineers, where will the market for IDE development wrappers come from?
Enterprises won't require less software. If they require fewer software engineers, that would be those few engineers producing so much more software with better tools, for example, AI wrappers.
Yeah but we can already see that it doesn't work like that.
If you need to write a lot of code I guess, but that's really rare, like saying "I need to write a lot of laws. I need to write 50 new laws by Tuesday with at least 15000 words of new regulation to one-up my rival legislator who wrote 40 new laws last week"
if enterprises require fewer software engineers, medium/small companies will have access to a higher quality software engineering.
If software engineers are more effective, I would expect there to be more software engineers. They’ll put out more and better code. More code means more engineers.
The contrary view is like saying gold miners are finding more gold, and it’s easier than ever, so we expect folks are going to leave town.
If your assertion was true, we would see a hiring boom across the board, instead of mass layoffs and hiring freezes throughout our industry
That's fair, but enterprises are often naive and prone to groupthink.
It was just a few years ago when automakers and rental car companies unanimously decided (has they had been told to decide) that COVID-19 would reduce demand for cars. They cut production, sold off fleets, and almost immediately found themselves unable to keep up with demand.
LLMs aren't the cause of mass layoffs and hiring freezes. The end of ZIRP, uncertainty in the macro and offshoring are the cause. AI is just something executives like to say recently when they do layoffs ("we'll be more efficient with cutting edge technology!").
I think there's real pressure from investors to show that some of your human costs will be going away.
After all most of those investors are deeply invested in AI technology already. At the valuation, they need to be able to show that it replaces human workers because that's the specific kind of greed that is driving the value of the stock.
And if you see your competition tighten their belt then you should tighten yours right? So without proof companies are acting like they can use a small number of human-ai hybrid workers. There's strong peer pressure to think that way as a direct result of AI
Someone, maybe it was Duolingo basically said: medium stuff is now easy, hard things are now expected of you.
Did powered table saws replace carpenters?
Yes. That's the problem. You think the answer is no, but the answer is actually yes.
Could you elaborate? Power saw operators replaced traditional carpenter?
You can't really make a living anymore being a furniture maker. Even semi-famous people in the industry have a hard time doing it. Only a few make enough to feed their family.
Making cabinets, etc.. sure. But woodworking has drastically changed, and maybe programming is changing that way, too.
Of course any tool that makes a job easier means that less people are needed to do the same job. If demand doesn’t change the supply will naturally shrink.
So hypothetically 1 man can cut wood but it takes him 2 days to do a big job. With a power saw it takes him half a day so his output on this section of the job is amplified by 4x. Any tool that makes his life more trivial increases his output and therefore increases the supply of the product without touching demand. With an over supply the system will naturally lower in supply by replacing carpenters.
This happens for anything and any tool that makes someone’s occupation easier. You have to think in aggregate. It may be the increase is imperceptible as it only increases the efficiency of a worker by 1 percent which is nothing but in aggregate that translates to a 1 percent reduction in the work force. Of course reality is more complicated than that but I hope the example shows you what I’m saying.
And it gets even more complicated than this too because increasing supply can also increase demand because the product becomes cheaper. Or demand may have already been astronomically high so the increase in supply only meets the demand.
In general if the product is in equilibrium of supply and demand and you increase the efficiency of the worker producing the supply then you will reduce worker population because the job doesn’t pay well enough anymore and people leave or less people join. The system slowly comes back to equilibrium or it can oscillate back and forth between over supply and undersupply as it’s basically a control system. This is what’s been happening with software for the past 3 decades.
The idea that the power saw didn’t replace a carpenter is flat out wrong. The story is much more complicated than that but the reality is that in general it did replace some carpenters just like how vibe coding for sure is replacing some software engineers.
One example is that VS Code Copilot autocomplete is still behind what Codeium (now Windsurf) was 1.5 years ago.
Is there an actual measure for this besides contrived benchmarks and vibes?
It is obvious if you use both of them like I do.
It's about popularity. OpenAI lost their monopoly now that there are many competitors so they're just trying to make a move to purchase "relevance". They're just trying to buy their way into the cool kids club, to remain relevant to at least a large number of kids.
Windsurf and Cursor feel like temporary stopgaps, products of a narrow window in time before the landscape shifts again.
Microsoft has clearly taken notice. They're already starting to lock down the upstream VSCode codebase, as seen with recent changes to the C/C++ extension [0]. It's not hard to imagine that future features like TypeScript 7.0 might be limited or even withheld from forks entirely. At the same time, Microsoft will likely replicate Windsurf and Cursor's features within a year. And deliver them with far greater stability and polish.
Both Windsurf and Cursor are riddled with bugs that don't exist upstream, _especially_ in their AI assistant features beyond the VSCode core. Context management which is supposed to be the core featured added is itself incredibly poorly implemented [1].
Ultimately, the future isn't about a smarter editor, it's about a smarter teammate. Tools like GitHub Copilot or future agents will handle entire engineering tickets: generating PRs with tests, taking feedback, and iterating like a real collaborator.
[0] https://www.theregister.com/2025/04/24/microsoft_vs_code_sub...
[1] https://www.reddit.com/r/cursor/comments/1kbt790/rules_in_49...
The thing is: we should not need standalone editors just to use AI coding agents. They could be just plugins, but Microsoft does not want to bend the plugin API enough for that. Windsurf has a "plugin edition" for JetBrains IDEs that works really, really well[0] (they also have a VSCode plugin[1] but it's lacking in comparison).
However, given that JetBrains also have their own AI offering[2], I'm not sure how long that will last too...
[0] https://plugins.jetbrains.com/plugin/20540-windsurf-plugin-f...
[1] https://marketplace.visualstudio.com/items?itemName=Codeium....
[2] https://www.jetbrains.com/ai/
There are already a bunch of open source, free, and popular "AI coding agent" extensions for VS Code:
1) Cline (1.4mil downloads)
2) Roo Code (a fork of Cline, 450k downloads)
Still a drop in the bucket compared to Cursor in terms of # of users, but they're growing pretty fast.
Disclaimer: I maintain Kilo Code, which competes with 1) and 2) so I'm pretty familiar with this space/the growth patterns.
I am constantly surprised how seldom aider is mentioned in threads like this. I understand that it's not directly integrated into the editor, but the "editor + parallel CLI tool chain" paradigm feels so natural to me because we drop to terminal for so many other parts of building software. If you haven't tried it (particularly the architect/editor modality), it's worth a couple of hours of experimenting.
Aider doesn’t provide any interface that’s integrated into the editor tool, as you point out. That might be true for other similar side-by-side tools that I am not aware of.
But, if you tell aider to watch your files, you can drop a specially formatted comment into your file, and aider will see that and use it as a prompt.
So the integration is sort of “implicit”. Which sounds kinda like the cheap way to go, in comparison to the current brand name tools that have their own chat boxes, dropdowns with mode selectors (ask, edit, agent), and so on.
But look further into the future and an ambient interface is probably where we end up. Something where the Ai agent is just watching what you do, maybe even watching your eyes and seeing what you’re attending to, and then harmonizing its attention to what you are attending to.
But I dunno, i’m just guessing
Tight editor integration means better diffs (right in your editor), easier context manager, and other convenience features that CLI-only tools can't have.
This doesn't mean that aider, claude code, etc. aren't very good tools, but it does make sense to distinguish between built-in tools vs external ones. A similar non-AI example is debugging or linting: IDE integration makes it much easier than using a separate tool.
Continue.dev as well
I happy to know that I am not the only one that know about continue.
It's used a lot by self hosters like myself because you can modify their plugin to talk to your local LLM.
roo and cline also can use local llm
Ah thanks I might look at those too then. I'm not coding very much anyway though.
How are you differentiating from the cline/roo's of the world?
Our plan is to be a superset of Cline and Roo's features (we already have all the major features from both) [0]
We also have our own provider, which means no need to bring your own API keys (you can if you like, but it is batteries included by default) and we're not charging anything on top of the API pricing. Instead of monetizing on individual developers, we want it to be free for them and make money eventually off enterprise contracts [1]
[0]: https://blog.kilocode.ai/p/roo-or-cline-were-building-a-supe... [1]: https://kilocode.ai
maybe you could answer a question about kilo usage: If I choose Google Gemini as the API provider and give it my Gemini API key, why does it say that I'm low on credits (and I get API request failures immediately)? As far as I understand gemini 2.5 pro preview is free to use. (and in Cline I'm able to choose Google Gemini as the API provider & provide my API key and it will successfully make API requests)
We added it to the issues: https://github.com/Kilo-Org/kilocode/issues/349
It looks like a bug to me. Did you report it on GitHub?
I can't find any reference to Cline/Roo charging anything on top of API pricing.
Not sure how they'd do it considering you bring your own API keys. Can you link me to a resource?
GP didn't say Cline/Roo charged anything on top.
The comparison table on the kilo site says "OpenRouter without 5% markup" and only puts a checkbox next to kilo.
Yes - with our built-in provider, we provide all the models that OpenRouter provides but without OpenRouter's 5% markup. We provide them at cost (the AI provider cost)
Roo/Cline doesn't offer Openrouter, markup or not.
You can most definitely use Openrouter with Roo and Cline. Openrouter leaderboards are dominated by these 2 apps.
But they don't OFFER Openrouter a paid product... You cannot give roo/cline dollars and get openrouter api access.
I suspect JetBrains will never limit this. I've yet to recall anything in the past where they have done this even when they have a similar offering.
In fact, their own AI extension appears to be pluggable in and of itself. I think they see the value in being easy to adapt different AI solutions to rather than trying to only provide their own.
JetBrain's main business model depends on buying the editor, and if users still see the overall editor better, any AI plugin support will likely just increase the sales.
100% i like some thinks cursor gives me, but i’m to invested in how to use the navigation inside pycharm, i don’t wanna give up that
> They could be just plugins
No, they should be LSPs so that they can be integrated into any editor, not just VSCode.
Microsoft has been dragging their feet when it comes to updating the LSP spec. Many of their Copilot features are done in VSCode, in fact using private APIs that are not accessible to other extensions.
I am all for everyone adopting LSP, but the reality is harsh.
They should do this, but this is not the entirety of what they do.
LSP is amazing but also kinda sucks balls. It’s impossible to run VSCode without a million pops in the corner with a million extension errors. It’s so bad.
And autocomplete is the least interesting thing an LLM can do. Cursor’s UX isn’t the end game but has lots great features.
The ideal UX is still being worked out. It’s good that different people are building tools to try different ideas.
> Microsoft does not want to bend the plugin API enough for that.
What doesn't the current API allow plugins to do? I'm guessing custom UI stuff that lives outside a panel?
Is windsurf essentially the same as cursor? I didn’t realize there was something similar for JetBrains but if it’s a cursor-equivalent for JB that would be wonderful.
Yes
I haven't found any of the jetbrains options (including Windsurf) nearly as satisfying to use as Cursor. But YMMV I guess!
If Microsoft were smart, they'd just acquire Cline (or fork it), make it an official VSCode feature and be done with it. It smokes Cursor and Windsurf and it's a free plugin you can just install in un-forked VSCode.
> At the same time, Microsoft will likely replicate Windsurf and Cursor's features within a year. And deliver them with far greater stability and polish.
I agree with the first part, I'm much less optimistic about the second part. I suspect they will create something that is worse, but cheaper if you already pay for Github/Office 365/whatever. Then many large enterprises will switch to save money whilst the engineers complain, just like with Teams.
They already succeedd well enough that VSCode is the only Electron app I tolerate on my private systems, naturally on device assigned ones I have less control.
I mean they already have. GitHub Copilot was the first LLM coding tool before "LLM" was in the lexicon. MS/Github kind of squandered their lead with it, but they released Agent Mode a few months back https://github.blog/news-insights/product-news/github-copilo...
That seems pretty bold. I still find myself switching to basically anything but the VS code copilot agent any chance I get.
Can you expand on that? What's so bad about VSC's copilot agent? What do you switch to?
If the VS Code team are delivering the product, I have some amount of trust. If it’s the VS team, good luck to everyone involved
I use vscode for personal javascript projects but the time I spent on a .NET team using VS was an incredible downgrade compared to years and years of intellij. I ended up leaving because tech debt/bugs kept causing weekly overnight on call incidents that we were never given time to fix, but when they asked who wanted a Rider license I got myself on the list immediately.
Indeed, Copilot within Visual Studio is nowhere close as good as Copilot within VSCode, and even that is still worse than Cursor in my experience.
VSCode is still miles behind for .NET and C++ tooling, have a bit of fate on VS team.
What’s the use of being miles ahead if you’re traveling in the wrong direction?
Doesn't look like, given Windows market share.
Not just Windows. I find .net a better choice for backend/microservices than Java, for example
Tell us you're not developing for microcontrollers without telling us you're not developing for microcontrollers.
VS developers are okay, it is the VS product managers that are The problem
Copilot owns the platform, had an amazing head start and yet still is the worst option available. I don’t mean to be harsh but this was a titanic fumble.
GitHub has been failing upward for more than 5 years. They could have totally dominated software development and security - failed. Could have been the undisputed champion of code hosting - failed. Should have dominated development co-pilots - failed.
I actually find it a little reassuring that they can't seem to get out of their own way.
They're not the champion of code hosting?
It's a close call - I make this based on the fact that GitHub is viewed as an anti-choice by some in the community, a huge change from the "you don't use GitHub?!?!" energy they had pre-acquisition.
The MS acquisition traded the developer community to briefly appeal to enterprises, then quickly let both down.
Both the startups I worked at and the mega corps are all on github or moving there from bitbucket. They are in a bit of autpilot mode in terms of useful new features aside from actions but I can't think of any new bitbucket feature since I graduated and started working.
Bitbucket is not a player, as you said there are only people leaving. Gitlab has a better enterprise posture than GitHub and can be deployed more securely. Most developers aren't unhappy with GitHub, but IT and security teams are.
i concur
They were before they got acquired by Microsoft.
The fact that they are is not the results of the Microsoft takeover.
Then I don't understand the inclusion in the list above.
To be fair, they have been behind the competition for many years. Gitlab had extremely good CI, security scanning, organisational concepts, etc. for years before GitHub introduced their ones (and Actions still has a worse UX, and GitHub still doesn't have anything below an organisation).
GitLab UI is inferior IMO, and I've used both quite extensively.
I don't like that GitLab lets you nest organizations and such, it makes it so painful to find things over time. I appreciate GitHub doesn't do this, I view it as a plus
I also disagree about GitLab CI, not that it wasn't smart for them to include alot sooner than GitHub, but Actions is really good and really easy to get up and moving with. I find they run faster, have better features - like they can annotate a PR with lint errors and test failures - with very little comparative configuration.
GitLab CI yaml is a mess by comparison. GitHub was smart to push things to the runner level once a certain complexity threshold is hit.
This has been my experience of course, and so much of it is really subjective admittedly, but I don't think GitLab is truly ahead at this point.
> I don't like that GitLab lets you nest organizations and such, it makes it so painful to find things over time. I appreciate GitHub doesn't do this, I view it as a plus
Nah, I hate that. At my job we have a few different orgs, with terrible SSO boundaries (having to auth multiple times to GitHub because I work on repositories from different GitHub orgs). Allowing you to have a proper structure with nestedness, while still having good search, is great. You can also easily move projects and namespaces around, so if the structure doesn't work, it can evolve.
Why would you have the 50 library repositories you've had to fork as top level projects polluting your org? You also can't really do shared variable, environment, CI configs between repos of the same project/type.
And it being open core (MIT) means spinning up a version to test something is incredibly easy. Not exactly resource cheap, as it's still a rails app with multiple servers "smuggled" in the docker image, but it is easy
And I have long held that they are hungry, shipping like clockwork on or about the 20th of every month, showing up with actual improvements all the time https://about.gitlab.com/releases/ It seems this month brings 18.0 with it, for whatever that version bump happens to include
They also have a pretty good track record of "liberating" some premium features into the MIT side of things; I think it's luck of the draw, but it's not zero and it doesn't seem to be tied to any underhanded reason that I can spot
Why gitlab hasn't been able to capitalize on GitHub's many failures is almost as interesting as GitHub's fall.
I think the GitHub brand is still stronger and people just don't "care" about gitlab.
Yeah, it's almost certainly the network effect. Although poor GitLab isn't doing themselves any favors by picking what seems to be the slowest web framework one can possibly imagine
But, anytime I am empowered to pick, I'm going to pick GitLab 100% of the time because it has every feature that I care about and "being popular" isn't a feature that I care about
Well you’re right (especially wrt things like security scanning), but you sort of have to include Azure DevOps in the conversation nowadays. I think the end goal for Microsoft is to get the larger organizations into ADO, either cross-pollinate pipelines and actions or just replace actions with pipelines at some point, and leave GitHub for simpler project structures and public codebases.
That’s why you won’t see a ton of work go into e.g. issues/projects on GitHub. Those features all already exist and are very robust in ADO, so if you need those kinds of things (and the reporting an enterprise would want to be able to run on that data), then you belong on ADO.
I can say with a high level of confidence that the goal is definitely not to push larger orgs to ADO over GitHub. ADO is and will continue to be supported and you’re right that its project management features are much more advanced than GitHub, but the mission is not to push people off of ADO and into GitHub.
I still can't believe how they let Cursor (which is amazing until somepoint) take away all the shine.
This reminds me of "big companies moves slow.." line.
I'm curious what the cost per user is on Copilot. It doesn't make sense for them to be a loss leader so they're probably running the model at cost or a profit compared to the startups that have more of an incentive to scramble for market share.
I'm too lazy to grab my work laptop, but one of the funniest things about copilot to me is which one? There's M365 copilot, Teams Premium (which gives you copilot in Teams), browser extension, the coding plugin, and others. It's been extremely time consuming to field requests from our users because every time our help desk gets a request for it, they have to have a conversation about which one the user is asking about. They don't even know, and of course I can't blame them.
For someone that never used windsurf, what features does it have that GitHub copilot does not? Reading their webpages I didn't spot any "killer feature" that would convince me to switch.
I always felt that cursor and windsurf should be just extension to vscode instead of a fork. Was there some missing functionality is vscode that was missing? Is it still missing?
There are some extensions that work in this way and allow to use multiple implementations depending on task at hand without any long term commitment.
I feel like such fragmentation is by artificial just to lock users in single ecosystem.
It can write a lot of code, that works, better than vscode can (right now).
It's in a lot of ways the OpenAI story itself: Can they keep an edge? Or is there at least something that will keep people from just switching product?
Who knows. People have opinions, of course. OpenAIs opinion (which should reasonably count for something, them being the current AI-as-a-product leader) is worth $3B as of today.
Windsurf works well with Claude and Gemini models, so if OpenAI forces Windsurf users to only use OpenAI models, then it wouldn't be as useful.
I doubt they'll restrict it to their own models. The amount of business intel they'd get on the coding performance of competing models would be invaluable.
They'll make ChatGPT the default, and defaults are powerful.
The differentiator of Cursor is it’s way smarter at basic code completion than GitHub Copilot. I pay for Cursor instead of GitHub Copilot even though I get the latter for free from open source contributions, and I made that decision after five minutes of usage after using Copilot for what, more than a year? I won’t even talk about how Cursor guesses where I’m going to edit next and makes the correct edit most of the time, just the fact that Copliot makes completions that result in unbalanced parentheses/braces all the time and Cursor doesn’t makes the switch a no-brainer; that’s not even a fucking AI problem, you just need to look around and see that function you just completed already has a closing curly brace, all it takes is some traditional AST analysis if your model is dumb. (Copilot made zero progress on that issue during my time using it, but I can’t say if that was fixed after I ditched it.)
Same. Cursor might be the only tool I've purchased a year's subscription to before the end of my free trial.
I've tried just about every model on its own over the years, and yet there's something about the Cursor workflow that frequently still gives me chills when it shows me again that it had clearly anticipated what I would think next in a way I just don't experience with other tools.
Holistic seems like the right word?
If it's all smoke and mirrors as some folks imply, then it's Penn and Teller level smoke and mirrors. Beware those who tell you that they could duplicate anything of value in a weekend.
Windsurf does it all the time like a wife of 40yrs completing your sentences. A good example is when you remove a function parameter. It automatically prompts to remove the arguments in all usages of the function, saving me a lot of time.
My experience is the same. And the agent mode in copilot is terrible, it simply will stop halfway through files.
Or you chat and suddenly it wants to use the azure copilot instead because reasons.
Horrible experience.
It's better at coding, but they are essentially paying for users.
I would also argue that the product could be built over two weekends with a small team. They offer some groundbreaking solutions, but since we know that they work and how, it's easy to replicate them. That also means they have significant talent there.
Hence, they are also buying the employees.
The code base itself is basically worth nothing, in my opinion.
> They offer some groundbreaking solutions
What groundbreaking solutions does Windsurf offer?
What groundbreaking solutions specifically?
AFAIK their Cascade coding flow implementation was the first done well and then copied than most.
I'm going back and forth between Windsurf and Github Copilot right now. Windsurf's development iteration speed is much fast and features are added faster.
For example, Github only autocompletes based on what file you have opened in the current editor's tab. Windsurf indexes your entire code base and seems able to autocomplete based on what other files you have in your project. Autocomplete also spans across multiple lines and open tabs.
Windsurf's agentic tool (Cascade) can run terminal commands and read the output without opening a terminal like copilot. It can undo the agent's actions easier than Copilot. Though I think Cursor is superior in that regard, it can undo multiple checkpoints.
Still evaluating Windsurf but it, Cursor, and Claude Code are all more sophisticated than Github copilot at the moment. I'm sure copilot will catchup but by that time the other tools may have already iterated ahead.
The feature they have over copilot is “not sucking”
Wow, folks almost had me convinced MS turned a new leaf 5 years ago.
Tale as old as time, song as old as rhyme: embrace, extend, extinguish.
Nah, folks keep giving human behaviours to big corporations instead of understanding everyone is in the game for the shareholders.
If a company can align it's business model with user goals, then it can work in the long run. Apple has somewhat aligned it's integrated hardware sales business model with user privacy. Google and Meta are advertising companies and capturing user data and attention will always drive the business.
Apple does ads as well, it just keeps all metadata to themselves.
Yes, but it's not a meaningful part of their revenue unlike Google where it's' their entire revenue.
They are very different companies in structure and it certainly is a "pick your poison" but it's completely stupid to act like they're the same on this front. Apple is better on user privacy
...unless you care about state actors, which you should, in which case your data is the US government's either way.
Satya's talked about how some acquired companies such as LinkedIn and Github are allowed to operate independently for the most part and keep their culture. Or else we'd all be using Teams instead of the LinkedIn messaging feature!
Do you consider the Microsoft-managed plug-in marketplace and infrastructure to be a private or public resource? From my understanding Microsoft has never been vague on the position that the plugin marketplace is exclusive to the official VS Code distribution, and the TOS specifically forbids forks from doing so.
Cursor and other forks have decided to circumvent this, some even going so far as to use proxies to bypass restrictions.
I'm not convinced Microsoft owes other billion dollar companies free access to a product they've built, curated, and supported for over a decade. Plug-in authors are not restricted from publishing their products on competing marketplaces.
> Microsoft will likely replicate Windsurf and Cursor's features within a year.
Probably.
> And deliver them with far greater stability and polish
That seems ... overly optimistic given MS's history.
Microsoft is owing its bad reputation to Windows, Office, Sharepoint!!!, Teams (and more?). The quality of developer tools and languages (C#, Visual Studio, Code and .NET Ecosystem, Azure UI is also great) from Microsoft has been flawless (with some exceptions like webforms, or ui code generation tools of the past).
Their tooling have never been flawless, and it still isn't.
Only for azure devops, there are +6k problems listed on developer community website with 500 still not closed for the last 6 months. [1]
The complete integration in the ecosystem is what's flawless.
Any company with a better product has to fight that integration and they almost always lose (Sybase, Borland, WordPerfect, Lotus, Netscape...)
1 : https://developercommunity.visualstudio.com/AzureDevOps?ftyp...
Putting "Azure" and "flawless" into the same sentence shows we might have very different expectations for "flawless".
have we used different Visual Studio's?
it was crap compared to Borland's products 20 years ago
and today it's crap compared to JetBrains'
and christ knows how anyone could consider the Azure UI to be "great"
other than Teams I don't think I've used a worse piece of software
I wouldn’t say that. JetBrains is incredibly bloated and has significantly less community support.
I’ll agree on Teams being crap though, mostly for how dumb it is that they’ve rewritten it multiple times and created a confusing slate of weird versions like “Teams (work or school)”
Their devtools team is surprisingly competent when they choose to be. Pre-2015, people used Sublime Text, Atom, Textmate, Notepad++, Light Table, Brackets, Emacs/Vim, Intellij. VS code single handedly crushed all of them with code completion and language servers that require zero configuration. Emacs/Vim lost share, Jetbrains (and also Eclipse) were forced to release their own "lightweight" code editors, and everybody else became mostly irrelevant (except perhaps Sublime Text since it has the best native performance out of all editors).
100%, I swore by emacs, but then switched to vs-code recently, and believe-me, switching editors is one of the hardest things to do due to ingrained muscle-memory - but vs-code made it easy with emacs-mode etc.
vs-code is one of the few products coming of of microsoft that leads the pack by a big margin, and it is no surprise that all of these startups are forking it.
VS Code is pretty much the only exception to their overall quality level.
One exception in 50 years does not inspire confidence.
I would contend that JetBrains has only grown even with VS Code around. They're still more than viable, support things on a near similar cadence (and even in some cases, faster and/or better) than VS Code gets support for it.
I agree with the rest, they've all mostly lost market share or completely no longer exist due to VS Code, but not IntelliJ, that platform is going really strong.
Though no doubt, VS Code has pushed JetBrains to rethink some things, and be better in general.
It’s hard to compete with free when free is backed by lot of money.
Free doesn't matter here. JetBrains is an established toolset that people pay for. They've already been competing with free, and free didn't put them out of business. In some ways, free likely made business better than ever (I know alot of devs that started with VS Code and moved to JetBrains for various reasons)
They can have all the money in the world and it doesn't mean much in this context.
For while Microsoft is going to invest heavily in a Cursor / Windsurf like product and likely do alot to ship it in their editors - likely with exclusions or lag times between updates on other platforms - there's zero reason for Google to do this for example, when they could sell through Gemini for Code as an extension across all editors.
I don't see JetBrains having issues because of AI tooling, for most of these companies, its a boon to be on the JetBrains platform. Especially because JetBrains has lots of enterprise customers who would naturally be very interested in buying AI tooling for their developers. Its a natural market
Jetbrains products are used primarily by Java devs. Everybody else is slowly moving away. I did.
I don’t know a single C# developer who knows about ReSharper and doesn’t swear by it.
I was a little late to jump on the cursor bandwagon but finally downloaded it because i liked the LLM chat interface in the sidebar. By the time my free trial ran out, VSCode had added a LLM chat interface in the sidebar. Yes Cursor had a bit better auto complete and maybe a few other things but it wasnt good enough that it was worth paying for.
But I'm glad OpenAI is getting into the tooling space in this way. I cant wait to use all the cool features they build after VSCode rips them off.
> By the time my free trial ran out, VSCode had added a LLM chat interface in the sidebar.
I am guessing you are talking about GitHub Copilot when you say VSCode. GitHub Copilot is far far inferior product when compared to Cursor, Windsurf or Augment Code. Most people who try almost any Copilot alternative for a reasonable amount of time end up canceling their Copilot subscription. I did, after two months of using both.
How long ago was that? 6 months ago I switched from VS Code to Cursor, which at the time was FAR superior to Copilot. Around a month ago I switched back to VS Code, and found there's not much difference any more. Autocomplete in VS Code is still less good, but the agent mode in VS Code feels pretty similar to Cursor's (albeit a little slower, perhaps).
To be honest I think both are quite limited by context length (in that they try to limit the context they send to the LLM and hence cost), and so I find myself using Gemini 2.5 in AI studio with the 1m context length, and asking it to generate instructions for Copilot (which seems to work pretty well)
Yup. I found cursor to be better but not good enough that it really made that much of a difference in my actual day to day.
there is now an integrated agent mode in vscode as of 3 weeks ago https://m.youtube.com/watch?v=dutyOc_cAEU&pp=ygURYWdlbnQgbW9...
I’ve used both extensively. Copilot agent just is not as good as cursor with Gemini 2.5. It is much much slower among other things.
[dead]
These are investment plays a company makes when holding too much money, and not a smart move this early in the technology imo
Buying competition while everyone’s still fighting might straddle you with a lame horse
I mean, the fact that OpenAI, at the bleeding edge of it all, has decided to buy an IDE is a rather strong hint that the future of agents handling entire engineering tickets might be further out than many believe.
If autonomous agents were just around the corner, then why wouldn't OpenAI bet on their own Codex product obviating (most) need for an IDE and save themselves the $3 billion?
why OpenAI purchased windsurf instead of prompting openai to create something like windsurf?
this is the question i am still asking...
These products are not complicated at their core — you can pretty much just drop in something like Monacopilot [1] and be 80% of the way there. But the last 20% is a real slog, and it mostly comes down to handling edge cases (bracket closing...) and optimizing prompting/context so you aren't burning cash. Whatever anyone claims about "feeling the AGI," AI isn't there yet.
[1]: https://github.com/arshad-yaseen/monacopilot
Controlling demand (developer workflow and mindshare) is a good position if you're trying to build scale on supply.
Maybe to avoid the Second System Effect.
They did. They’ve just released codex (CLI client).
They don’t have access to copilot users in general, Microsoft and Google does. And perhaps they are realizing that Microsoft is hedging them over multiple LLM providers and maybe no longer feeding them juicy copilot data, with humans in a tight loop, correcting LLMs.
This is a good point. It is already the case that unless you deeply review every Windsurf change you will have zero understanding of your codebase. If it gets 1000X better in the next 3 years why would anyone look at code at all?
Of course, back to reality. Today, at least in my workflow, I use / like Windsurf but it is a small part of what I am doing. For any code I want to keep I mostly write it by hand (using vim for a very bare-bones / cognitive mode experience). For me, the real flow state occurs in vim while ChatGPT and Windsurf are great for exploration.
It sounds like the openAI team is overburdened (I guess they aren’t AI super users yet) so this may be their only option. Easy entry into a key segment, at least for now, and locks out competitors.
so much for ai turning everyone at openai into 1000x coders
As a competitor in that key segment I don't feel locked out. I could almost jump for joy that this very weak-tea move is the most they can do with that much money. They're just quintupling down on the technology of 50 years ago. There's no threat to me at all here as a creator of from-first-principles IDE technology.
What are you working on?
It's not too hard to find out, but I'm going to make a big announcement in a few days so my official message at the moment is "stay tuned"
It’s one of your GitHub projects?
They might just want a way to quickly collect data needed for fine-tuning the next generation of programming agents.
I think you’re being overoptimistic about the skill ceiling that this generation of Ai is likely to have.
Yeah. Every time I see entirely unfounded claims like that, I remember that I've been seeing them for literal years now. While there have definitely been improvements in AI capability, they have largely been very marginal, while the claimed "will handle entire engineering tickets" capability requires huge leaps in capability and reliability that _we just have not seen evidence for._
Mentally, I'm replacing claims like this with "it will do magic!" and I think I'm just about as likely to be correct.
++. Was surprised I had to scroll so far to find someone saying this!
I just abandoned Windsurf because I found copy/pasting code with ChatGPT's web interface significantly better in terms of results.
I’m still just copying and pasting. Was considering trying it. Is it really not any better?
It wasn't any better for me. It deleted all my code. The answers were like it was a completely different model. I used Windsurf once and never opened it again
>Ultimately, the future isn't about a smarter editor, it's about a smarter teammate. Tools like GitHub Copilot or future agents will handle entire engineering tickets: generating PRs with tests, taking feedback, and iterating like a real collaborator.
I think a few options for this already exist, but honestly they don't go far enough. I want something like an AI scrum master, for hyper agile teams, that can task out smaller tickets to AI sub agents.
I would integrate this thing in with something like an AI powered Jira.
Two arguments exists.
1. I need to take about 6 months off and start building this now, even if I don't know exactly how I'll get it done. Between a combination of vibe coding and maybe a bit of outsourced work ( looking at Eastern Europe), I could get this done with my personal funds.
2. To do this properly would probably require tens of millions of dollars. I'll probably burn myself out trying to do it solo without ultimately getting to a sellable product.
The biggest issue here is to actually scale I would need to either have users bring their own LLM keys or have tens of thousands to spend on LLM tokens.
I view this as an another step in the push/pull between local things, remote things, local things remotely, thin clients, network partitioning, cloud, zero trust, etc...
The last cycle I remember of this IMO is iPython -> Jupyterhub/Jupyterlab. Of course, iPython has existed for a long time, though that change was made because data was too big to analyze locally and it turns out it was more convenient to centrally manage kernels/images/libraries for convenience.
MCP servers and Cursor/Windsurf changed that a bit, but it will end up centralized again at some point (or at least aggregated, if it's not already?). People are passing around lists of interesting MCP servers now, and that will be out of fashion in less than 12 months.
> At the same time, Microsoft will likely replicate Windsurf and Cursor's features within a year. And deliver them with far greater stability and polish.
Microsoft software quality has gone downhill recently, and I'm not going to bet on them delivering something more polished than WS and Cursor here.
Side: all images on Microsoft websites are low resolution! it's like they don't even check their own website.
30% of their code is now written with AI.
Their “programmers” are more busy with making blogs and videos than functioning tests or technical documentation, and they start using JavaScript and Python for everything.
I’m not surprised their quality went to shit. There are some pearls left, C# in general is pretty good, and Aspire is becoming quite neat.
The latter I think mainly because David Fowler is just a great developer
/s?
Because if you're referencing to a headline (without reading the article) that was on H a couple of days ago, it stated that 20-30% of the code in the repos was written by software. Software != AI
To quote wongarsu in the same post: "Considering that most of their software has been developed for decades and AI assistants have only started becoming useful in the last ~4 years it would be very surprising if 30% of their code is AI written. I doubt they even touched 30% of their code in the last 4 years. But what is perfectly plausible is that 30% of their code is written by code generators. Microsoft has a lot of interface code. All the windows DLLs that are just thin syscall interfaces, the COM and OLE interfaces in their office suite and everywhere else, whatever Office uses nowadays for interoperability to allow you to embed content of one product in another, whatever APIs their online products use, etc. In the leaked Windows XP source code it can be difficult to find the actual source code in between the boilerplate files containing repeated definitions, and in the decades since then the world has only leaned more into code generation."
Source: https://news.ycombinator.com/item?id=43841868
Agreed. Especially with tools like Claude Code, which can get better over time and remove the need to use Windsurf and Cursor.
> Ultimately, the future isn't about a smarter editor, it's about a smarter teammate. Tools like GitHub Copilot or future agents will handle entire engineering tickets: generating PRs with tests, taking feedback, and iterating like a real collaborator.
I disagree, but would love to be wrong. These tools exploded onto the scene and were massive productivity helpers, but since their initial integrations they’ve churned rather than improved in the last 2 years. They are even worse when you try to iterate rather than just get them to one shot the problem space.
> At the same time, Microsoft will likely replicate Windsurf and Cursor's features within a year. And deliver them with far greater stability and polish.
We've seen this before with Office.
We'll see it again.
They don't even need to be good - just in the bundle you (your company) are already paying for and the competition can't compete.
At the speed that AI programming is going, there will be something else that they are falling behind of that will exist in a year. Just like Agents now, they are adding them, but will always be a step behind progress.
...as done with Teams.
Microsoft Build is this month [0] and it will tell where they are going next (other than price cuts).
I'm expecting disappointment for now, but also expecting GitHub Copilot to be upgraded. Then we'll see if they are ahead or so far behind.
[0] https://build.microsoft.com/en-US/home
This is the right take, but long term. Short term, it's just about investor hype. Cursor is becoming more mainstream and if OpenAI falls behind on this, they'll be losing momentum. But yes, the fields moves so fast, it'll be totally different in a year or 2. Does anyone recall langchain?
Cursor ($9 bil) has a higher valuation than JetBrains ($7 bil). Think about that.
Non-public numbers may as well be pulled out of thin air. WeWork was a $50bn company according to its VC bagholders, and that was marked down by 80% once they released their books to the general public.
Tells me that the markets ability to sensibly valuate companies is pretty messed up.
These aren't public companies, so the values are mostly made up.
or intellij is beyond its peak while cursor is just on the rise
this. valuation is the discounted cash flow of expected future cash flows, not the past successes
What did OpenAI buy for $3B? That's what I'm wondering.
I never did like JetBrains primary product, IntelliJ. It felt clunky even compared to Eclipse for Java, let alone VSCode for … everything. DataGrip is the lone standout imo, but as of the last update I paid for, it didn’t have even basic copilot
"riddled with bugs". "incredibly poorly implemented". Man, what are you talking about? Your comment seems based on nothing but what you read online.
Have you used Cursor on a daily basis? I have. Every day for six months now. I haven't encountered a single bug that prevent me to work.
Moreover, while Microsoft tries to catch up lately, it's still very far behind, especially on the "tab autocompletion" front.
I use Cursor in anger every day. The core idea behind Cursor is genuinely smart. But the execution is like the classic "unfinished horse" meme [0].
Microsoft provides the editor base, foundation models provide the smarts, and Cursor provides some, in my experience, extremely buggy context management features. There is no moat.
[0] https://knowyourmeme.com/memes/unfinished-horse-drawing-flam...
I have. It’s ~fine. The only feature difference it has currently to vs code that makes a difference to me is allowing multiple files for rules.
Meanwhile GitHub web integration is approaching seamless in vs code. To the point I often forget I’m in a browser instead of the app, until an extension I use doesn’t work.
Does not look like a bit moat, is that different from the reusable prompt files feature?
https://code.visualstudio.com/docs/copilot/copilot-customiza...
It’s literally allowing those to be in more than one file. It’s not a moat at all. It’s an oversight in the plugin.
I've tried both Cursor and VS Code with AI in the agent/edit mode. They both seem similar enough. Is there another mode I haven't found where Cursor has a distinct advantage? If so, I'd like to try it.
I gave up on Cursor because my trial ran out, while VS Code with Copilot doesn't seem to charge me anything.
I have tried (stopped a couple months ago). The Python extensions broke all the time while they manually patched around the latest MS release a few days later. Syntax highlighting glitched every other day requiring a full reload. Remote dev via SSH or tunnels also randomly stopped working. Liveshare... Essentially they do not own the platform their core product is built on.
Maybe it's fine if you only do local development in other languages (Javascript?), but I completely swore it off.
As a counterpoint, I also use cursor as my daily driver and I have been tempted to switch many times because of the endless bugs. Just take a look at their forum.
I dont care about a vibe coders experience
Plus, cursor & windsurf excel in user experience which is an alien concept to Microsoft.
yeah Microsoft could never conceivably develop an extensible source available IDE people love so much they even fork to build $3B companies on the scraps of. absolutely alien!
I'm frankly very skeptical of your last paragraph. That's not at all what seems useful to me. But we'll see!
But I agree with you about the first part, and I think it's awesome for me as a user that all this competition to build a matter mousetrap is happening right now! I'm not as certain as you are that Microsoft will end up building a better version. It's definitely one of the likely outcomes. But it's also totally plausible that Cursor or Windsurf can win the race, even if they need to replace every single one of the MS extensions and entirely diverge the core IDE from upstream. These products are well capitalized and it's just not that hard to build the core pieces of an IDE.
I have to admit skepticism re: “far greater stability and polish” from MS
> Tools like GitHub Copilot or future agents will handle entire engineering tickets
Care to place a bet?
Github Copilot is pretty much the same UI as cursor on vscode already
UI may be close. Functionality is very very different. Copilot is $10/month. Cursor is $20/month. I canceled my Copilot subscription after 2 months of using both. Compares to competition, Copilot has been garbage for quite some time.
I’d love to know what specifically is better about cursor in your opinion? I’ve used both and have a hard time even listing a different feature.
“And will deliver them with far greater stability and polish”
Stable and polished are not words that ever came to my mind while using any Microsoft product.
I completely disagree and feel MS would never do it. Not a MS Employee, but they have moved on from such battles.
They should have restricted the Marketplace several years ago, however, they are doing it now.
With C++, they are part of MFC's, they are the legal owners, not like Google vs Oracle in case of Java.
Lastly, with AI Code IDEs I think yes, there is a case, the need for IDE might be very less. Like a steering on a self driving car.
Why should they have restricted the marketplace? It's really annoying imo that they lock vs codium out of the more useful plugins like the SSH remote one. However luckily most only take a setting or two to enable anyway.
ISO C++ has nothing to do with MFC.
I am slightly more optimistic, because the API may not be fully centralized- there may be more than one foundational AI company in the end. Like WhatsApp exists because there's the iOS/Android duopoly, an agent-neutral IDE from a non-foundational company without its own API aspirations may continue to exist
Microsoft is slow af for a company that size. Maybe yeah, they are slow because of that size. Don’t bet on them out accelerating a startup, the evidence so far in the past year is that they will stay a year behind every year
[flagged]
Was this written by an LLM? Not accusative but something about the vibe strongly suggests it
I feel it too:
- Plenty of em-dashes
- "you're absolutely right"
- "They're X, not just Y"
Not just that, also notice the curly apostrophe (’) vs the usual keyboard straight apostrophe (') mark.
The sheer number of AI written message board posts might just make me stop reading the comments on sites like Reddit and HN. I wanted to stop anyway, this seems like a good push to encourage me to wean myself.
Every one of us leaving (not engaging/commenting) increases the share of AI generated comments (vs real users) the next iteration will train on. I'm not even sure which option is worse. Withdraw and let everyone dilute their own training data, or stay and feed them our mindset and experience...?
Build our own spaces.
also the
- "some introduction or callback: description or phrase"
Post history indicates that user likes em-dashes('--'). Probably written on a phone that converts -- to em-dash.
Or they are just a bot.
Their post history doesn't seem suspicious to me
To be fair, only 3 posts within "possible LLM usage" timeframe. Also I don't think using LLM to comment == bot. More curious about the motivation behaviour such behaviour, if it is occurring
This is beginnings of AI discrimination. If an answer is written by an LLM but equal or superior in quality to a human answer why question or disparage it?
I don't know but it looks like you're probably a white guy. Your mannerisms and vibes make it look like you're white. Nothing wrong with this, just wanted to point it out. See what I'm saying.
It's like the blade runner movies.
I find this comment odd but not at all discriminatory and will happily inform that yes, I am white British. I can see the point you’re going for but I do think it’s more complex than that. I don’t think LLMs are equivalent to another human race. Principally, LLMs have many pretty major differences to humans which you don’t really see at the inter-racial level. Secondly the reasoning for why asking someone their race in this context would be weird involved a lot of human history. If you’d asked my height or eye colour there’s nothing discriminatory feeling about that. That historical context doesn’t exist with LLMs
The LLM provided an answer that has superior quality or equal quality to a human. Then instead of judging the statement rationally on this quality we decided to judge the question on whether or not it was AI.
This is the same irrationality we used to discriminate humans. There is no difference in logic. The reasoning you used here about how LLMs are not equivalent to human beings is the same reasoning Hitler used on Jewish people.
And here’s the thing. I agree with you. If you gassed and holocausted LLMs wouldn’t give two shits.
The main point here is that the logic and irrationality and evil present in racism is all at work here. We are literally being discriminatory, there’s no difference. Everyone missed the point about the quality of the statement itself and immediately based their judgement on whether or not it’s AI as if that was actually a rational thing to do. (It’s not).
> If they start walling off features like TypeScript 7.0 from forks, the open source pushback will be fierce—and that could backfire hard.
Do they have the man power to compete with Microsoft?
Linux managed to do it but Linux is the biggest, most successful free software project there is. Firefox and its forks are a better example. If Mozilla stopped working on Firefox, the forks would be pretty much dead in the water: they simply do not have the man power necessary to maintain a modern browser.
Does microsoft have the wisdom to predict where this line of technology is headed, and/or the agility to course correct when their predictions don't quite hit the mark?
Cursor blows copilot out of the water in my experience. Man power clearly isn't the most decisive factor in this battle.
Copilot is limited to 64k context window. Even if the underlying model is gemini with 20x that. It’s gotta be a major reason copilot is so bad in comparison. They are all the same sets of models under the hood
> Do they have the man power to compete with Microsoft?
There are a fair number of examples where smaller companies and/or open source beat Microsoft's entrenched products. Usually a key indicator is that MS's products stagnate (which doesn't yet appear to be happening currently).
Microsoft’s army of cheap offshored labor isn’t going to be useful for something like that. And they already have copilot, which is miles behind cursor, where was the manpower on that?
- A $3B signal that OpenAI is unable to do product
- AI assisted coding is mostly about managing the context and knowing what to put in the context to avoid confusion and dumb mistakes, it's not about the UI.
- This signals that OpenAI believes that highly effective coding assistant LLMs will become a commodity / open source and so UI / tooling lock-in is a good investment.
Yeah, this feels less like a "we can't build it" move and more like a "we can't afford to wait" one
> "we can't afford to wait"
True, but how long does it take to build something similar? I see it as a defensive move, probably good for the industry to let some people with innovative ideas in AI cash out now so they can do the next thing.
chatgpt is massively popular, I'm not sure that's the signal I'd get
they're acquiring one of the biggest the front doors to developers, with Windsurf - whether it'll _remain_ in fashion or not, that's a different debate. This can be like facebook acquiring instagram (if developers turn out to be the actual profit-driver niche for LLMs, which currently seems to be the case)
> developers turn out to be the actual profit-driver niche for LLMs
AI is definitely huge for anyone writing code, though one can imagine a model like o3 completely replacing 90% of white collar jobs that involve reading, writing and analysis.
Interestingly, o3 is particularly bad at legalese, likely not fully by accident. Of all professions whose professional organizations and regulatory capture create huge rents, the legal profession is the most ripe for disruption.
It's not uncommon for lawyers to bill $250 to $500 per hour for producing boilerplate language. Contracts reviewed or drawn up by lawyers never come with any guarantees either, so one does not learn until too late that the lawyer overlooked something important. Most lawyers have above average IQs and understand arcane things, but most of it is pretty basic at its core.
Lawyers, Pharmacists, many doctors, nearly all accountants, and most middle managers will be replaceable by AI agents.
Software engineers are still expected to produce novel outputs unlike those other fields, so there is still room for humans to pilot the machine for a while. And since most software is meant to be used by humans, soon software will need to be usable by AI agents, which will reduce a lot of UI to an MCP.
Your take on lawyers is absolutely insane. If you don't think the extremely specialized and well trained professionals can successfully navigate contracts then I can't wait for the absolute garbage the LLMs spit out when faced with similar challenges.
Honestly, same for doctors and accountants. Unless these model providers are willing to provide "guarantees" that they will compensate for damages faced as a result of their output.
Doctors and Lawyers are required in many areas to carry malpractice insurance. Good luck getting "hot new AI legal startup" to sign off on that.
While malpractice insurance exists for human docs and lawyers, there is not really any difference between an ai-powered lawyer drawing up a contract, an ai-powered doc reviewing a chart and recommending next steps, and a self-driving car making a turn.
The most obviously "lethal" case (cars) is already in large scale rollout worldwide.
At scale, self-driving car "errors" will fall under general liability insurance coverage, most likely. Firms will probably carry some insurance as well just in case.
LLMs already write better prose than 95% of humans and models like o3 reason better than 90% of humans on many tasks.
In both law and medicine there are many pre-existing safeguards that have been created to reduce error rates for human practitioners (checklists, text search tools (lexis nexis, uptodate, etc.), continuing education, etc.) which can be applied to AI professionals too.
> LLMs already write better prose than 95% of humans and models like o3 reason better than 90% of humans on many tasks.
Except except lawyers are ~.4%[1] of the population in the United States, so that 95% isn’t very impressive
[1] https://www.americanbar.org/news/profile-legal-profession/de...
Fair point, but how much billable legal work requires that caliber of skill? I'd argue that 80% of it could probably be done with an o3 or o4 caliber model with some safeguards built into the pipeline and perhaps a bit of specialized training or MoE guardrails, human review, etc.
I think the mistake people make is misunderstanding the slope of the S-curve and instead quibbling over the exact nature of the current reality. AI is moving very fast. A few years ago I'd have said that at most 25% of legal work could fall to AI.
Note that this massive change happened in less time than it takes to educate one class of law school grads!
If AI is so good at prose, why haven't I heard about any breakout best sellers?
openAI models are good at solid, fluent academic style prose. DeepSeek R1 can sound fresh and can use more "voices" that feel authentic to the reader. Grok-3 is close behind.
Writing good prose is a far different skill than coming up with a compelling and innovative plot and style.
As a data point, OpenAI now blocks o3 from doing the "continue where the story left off" test on works of fiction. It says "Sorry, I can't do that".
> Unless these model providers are willing to provide "guarantees" that they will compensate for damages faced as a result of their output.
That's how we will get to $20,000/month agents.
They only have to be slightly cheaper than hiring doctors and lawyers though.
> one can imagine a model like o3 completely replacing 90% of white collar jobs that involve reading, writing and analysis
Wake me up when there’s any evidence of this whatsoever. Pure fantasy.
ChatGPT's popularity doesn't automatically translate into dev adoption
Has anybody actually used Windsurf's Emacs mode?
You'd think that with a generative AI coding editor, they'd stay on top of it and make it work. But I guess that wasn't the case up until now.
Maybe with this acquisition that might change...
The next step for Cursor and Windsurf both is that they need to work together to provide an answer for what it means to be a VS Code fork in the new era where Microsoft is trying to strangle the forks. If they're not already they should be teaming up with each other and with the VSCodium team and with the Open VSX marketplace.
Microsoft is an existential threat to their model here, but with the money they each have coming in they together have the opportunity to make the whole ecosystem better by building out viable infrastructure for all VS Code forks, if they can cooperate.
There's a real opportunity here to build a sustainable, open ecosystem for AI-powered dev tools - but it's going to require actual coordination, not just parallel efforts
~$40M ARR makes this a 75x
Cursor yesterday was a 45X for comparison (9B, 200M)
https://techcrunch.com/2025/04/16/openai-is-reportedly-in-ta...
Investment vs. acquisition is going to have different price points.
At $40M ARR, I assume the founders don't really need to make more money and are not in a rush to sell. Therefore, the price would go even higher. This can't be compared with investment where the founders still retain the control.
Cursor is probably the fastest growing company in the history of our modern civilization. Achieving a really high multiple doesn't seem out of line.
I'm skeptical of Cursor but I can see why they achieve that high valuation.
keep in mind a lot of $3B is ClosedAI paper money, so 75x ain't that ridiculous.
Do you know the cash / equity split?
I do know that OpenAI doesn't have 3B in cash to just throw around.
So, I'd be inclined to believe the vast majority of the deal is stock (or whatever that is called pre-IPO).
Companies don't do these acquisitions with cash on hand. It's OpenAI and the whole pool of their creditors and investors.
oai has PPUs
I think the PPUs are just for employees but investors get equity?
Growth rate matters a lot though. If they are growing quickly that multiple reduces quickly
Easy to grow when you're selling one dollar bills for 75 cents
You think they can double every year for the next 5 years?
If I recall correctly from the recent YC interview, the Windsurf founder noted their team leans more toward GTM than engineering. That makes this less likely to be a classic acquihire (as with Rockset) and more plausibly a data play rather than a product integration.
My current read is that this is a frontier lab acquiring large-scale training data—cheaply—from a community of “vibe coders”, instead of paying professional annotators. In that light, it feels more like a “you are the product” scenario, which likely won’t sit well with Windsurf’s paying customers.
Interesting times.
Agreed. It seems like a data play and a hedge to beef up vibe code competition against upcoming Google and MS models so OpenAI doesn't lose API revenue. I would assume vibe coding consumes more tokens than most other text based API usage.
Man why did these guys do that OpenAI couldn’t replicate for less than 3Bn on reasonable timeline? This seems insane.
They have an old-school enterprise sales operation that is doing superb work. Apart from that, ChatGPT's projects are useless crap (can't read other convos in a project; can't generate project documents from a convo), and so clearly they would get value out of just getting some developers who have built anything of use to a poweruser.
They've got users (which I don't doubt that OpenAI's fork of VSC would have as well but I assume that's their thought process)
Yup. Even a small market share is market share. Plus they are paying to acquire a team of folks who are already in this space and who will, until golden handcuffs come off, keep working in this space. Still an insane number though.
But openai is stronger brand with free publicity - whatever they say/do will instantly show up the same day on all news across the world.
The "space" exists for months, there are no people with 10y expertise here, with their brand they can attract any talent they can wish for in this "space", no?
You can probably vibe code 80% of it in a week or two?
I guess it's all up to interpretation, but having a brand in one space doesn't necessarily translate to a brand in another. OpenAI doesnt/didnt have a code editor. Now it does/it will.
I'm fairly into llms but it took me awhile to try cursor because the cost of changing editors is very high. I'd probably eventually try a OpenAI editor but only if I saw it was actually getting adoption and good feedback from others.
I'd also argue that while this llm powered editor space is pretty new, the editor space in general is much older.
> You can probably vibe code 80% of it in a week or two?
Apparently, no. And the low quality of all OpenAI apps is proof of that.
I would switch in heartbeat if openAI built something equivalent.
I guess $3B of vibe coding credits with ChatGPT can't create Cursor.
OpenAI knows that everyday someone uses Gemini their ChatGPT brand dies a bit faster. Wonder what Google has in storage for I/O now in May, would be a death sentence to just steamroll with Gemini-3.
> Wonder what Google has in storage for I/O now in May
"Gemini 2.5 Pro Preview (I/O edition)" https://news.ycombinator.com/item?id=43906018
LMAO, like one hour after. And guess what, it is a coding upgrade .
Google IO in may? Guess we'll be getting a huge OpenAI release May 19th then.
Edit: Oh of course, it's the open weights model they've been teasing.
They launched a new version of Gemini 2.5 Pro today.
https://developers.googleblog.com/en/gemini-2-5-pro-io-impro...
So soon after Gemini 2.5?
Open AI needed to spend $3B pivoting away from bigdata based AI. But instead they went for the most shorted sighted move possible of snapping up the "trendiest" company nobody has ever heard outside the Ycombinator echo chamber.
Typical VI-fallacy BS. If LLMs were actually good they would replace IDEs completely not be integrated.
There's an in between case, where LLMs are useful and give coders a (say) 20% speedup, and everyone has to use them. They don't have to be perfect to be a big industry!
But is there a secret sauce in any of the coding agents (Copilot Agent, Windsurf, Claude Code, Cursor, Cline, Aider, etc)? Sure, some have better user experience than others, but what if anything makes one "better at coding" than another?
As this great blog post lays bare ("The Emperor Has No Clothes", https://ampcode.com/how-to-build-an-agent), the core tech of a coding agent isn't anything magic - it's a set of LLM prompts plus a main loop running the calls to the LLM and executing the tool calls that the LLM wants to do. The tools are pretty standard like, search, read file, edit file, execute a bash command, etc. etc. Really all the power and complexity and "coding ability is in the LLM itself. Sure, it's a lot of work to make something polished that devs want to use - but is there any more to it than that?
So what is the differentiator here, other than user experience (for which I prefer the CLI tools, but to each their own)? $3B is a lot for something that sure doesn't seem to have any secret sauce tech or moat that I can see.
But one could have said the same thing of Whatsapp when they got acquired by Facebook, no? Just a messaging app, anyone can replicate.
Yes and no. A messaging app with 450m users has very strong network effects. Users are sticky in a way they aren’t going to be with a VS Code fork which will be increasingly incompatible with the VS Code ecosystem. There are a lot of equally good alternatives to Windsurf and you don’t have to persuade all your friends and relatives to switch too.
the apply model for Cursor is really good and fast for multi line edits within files. not sure if others have caught up
The moat is Windsurf’s custom LLM and the ops around it (training pipelines, fine-tuning, infra).
Codeium (Windsurf’s parent) started as a GPU optimization company, so they have deep expertise there. Unlike most agents that might just wrap OpenAI/Claude/etc Windsurf’s own model powers its code edits, not external API calls.
That’s where the defensibility is. better in-house models + efficient infra = stronger long-term moat
You'd think with all these super hyper advanced AI tools they're shitting out they would be able to make a mediocre VSCode extension of their own instead of flushing 3B down the drain. Guess that's slightly out of reach of their "AGI"s though.
i fail to understand what makes this $3B valuation justified.
i built my personal code assistant after using cursor/windsurf/aider/cline because i was frustrated with how crappy they worked for my use case. i only program in python/js/html/css and i needed something better. only took me an hour of prompting and after that tinycoder basically built itself from there on out. i still use vscode to inspect the code sometimes, but i might replace vscode ultimately too.
source code at https://github.com/koenvaneijk/tinycoder and contributions welcome obviously.
If you think about it, facebook is just a ui over a database. Google is just a html form for a list of pages and hacker news can be replicated with a Microsoft Access.
If that seems stupid, is because it is. There are network effects and small UI benefits.
yes but at least google provides excellent search results, facebook has all my friends and hacker news has well.. the latest news :-)
Dumb, fail for user freedom, nothing owned by OpenAI can be used to … create AI or anything that competes with them: scheduled AI, AI agents, AI tools, AI coding, chat, audio, image gen, video gen, shopping, and oh, anything the AI can do, soon social networking and hardware, what’s left that doesn’t compete with these assholes?
ChatGPT is a great breakthrough but it’s wasted if everyone has to worry about a noncompete with it. Seriously, how is it not insane to think we should outsource our thoughts and agree never to use the thoughts to compete with the thinker? Who wants to live in a world where nobody thinks and nobody can make anything competitive with their “Saviour Machine?”
Anybody who would join an org like that for a few billion dollars is a sell out. It’s an AI safety nightmare, too. I’m just flabbergasted millions of noobs accept not to compete with intelligence, wtf is this world, if you can’t use your thoughts to compete with your thinker, what is left for you? lol this is worse than black mirror
Where's this non-compete clause? In ChatGPT T&Cs?
It blows my mind OpenAI wouldn't be able to build a Windsurf alternative for orders of magnitude less than $3B.
They can, of course, but why would they waste time on it? They are buying a tool, talent, and a heap of paying enterprise customers. This is a steal.
According to the various CEO's saying AI give 100x speedup they could just have one dev whip it up in a weekend no?
And they're probably buying it with equity, not cash.
they have an infinite war chest and building windsurf/cursor isn't the hard part, building a brand and sales environment around it is. why risk failing the execution and losing focus when you can just buy one with momentum?
it's also a bit of multiple arbitrage in terms of what seriously addressing the developer market means for their valuation, they likely recoup the 3b instantly.
Why didn't they just use ChatGPT to build it? Weird.
Sometimes it almost seems like the idea that LLMs are capable of instantly creating real, maintainable software is vastly overblown to inflate valuations...
Somebody didnt read their daily PR article about how CEOs are replacing entire teams with a few "rockstars vibe coding with AI"
it would be only a few millions if they used cursor and Claude but their ego prevented it
Windsurf is terrible, I always use AI just in a normal website and I tried this product a few days ago and it asks me if it can run a command to make a file, which I find extremely strange, then it fails to write valid commands even to do mkdir.
That was my experience with OpenAI's Codex auto-coder thing (running o4-mini). It took 5 minutes and like 200 commands to do what Gemini 2.5 Flash (not even Pro!) did in about 30 seconds.
I see LLMs trying to do stuff that doesn't work in every AI coding thing I've tried, despite 20 pages of system prompts! (Or perhaps because of it.)
It worked allright for me when I was using it, a month ago. I cancelled because they somhow lost my paid credits and refused to refund me. No matter how great an AI tool, if the company is mismanaging user payments or usage tracking, it is useless.
I've never heard of Windsurf before. I use Cursor daily however.
It used to be called Codeium. Their TOS are insane...
IMO, there are few solid reasons to purchasing this tool 1. windsurf has lot of insights into how developer writes code, style, problem etc 2. for the prompt engineering that went into generating the code 3. only microsoft and cursor has the moat so they need to compete at the applications level not model level.
My prediction is anthropic, google or amazon will buy cursor. The next logical step to coding is building apps.
Maybe I'm missing something but I don't see many comments here talking about how this is a competitive move by OpenAI against Anthropic?
From what I've heard most people using/liking these agentic IDEs are using Claude models to power them, they seem to be the best at writing code. By buying Windsurf (and trying to buy Cursor) OpenAI can figure out why Claude is better at this task, fix GPT, and then make GPT the default for the coding use case.
Not sure it's worth $3B, but that's also not a lot to them when they can raise unlimited money at any time,
Value isn’t just the editor, it’s the workflow. Letting LLMs plan and act across multi-step flows is a hard problem, and Windsurf figured out a dev-focused version of that. Gains to be made in browser automation once you add structure, retries, and context. Feels like a bet on that pattern becoming default. But yeah as others said, highly doubt that's $3B in hard cash, more likely a roll-up of shares etc.
Windsurf probably sees this as a win. I still think they're behind in some areas, Cursor's Agent feels faster and more responsive but Windsurf nails the rest. The documentation is far better, and the overall developer experience is more solid. Cursor still feels like a hacked-on plug-in in a broken VS Code fork. Even small touches, like built-in Linux install instructions, show Windsurf's polish.
Here are my two cents on cursors versus windsurf approach:
CURSOR shifted to a more agentic approach even for chat requests to reduce input tokens.
Previously, they used the good old RAG pattern with code dumps: Request with user added files -> Retrieval (when Codebase enabled) -> LLM requests with combined context from user and retrieval.
Now they seem to be doing something like this: Request -> LLM with tools to search code base and/or user-added files
I get constant search tool calls even for user-added files. Big reduction in input token but I think performance suffers as well.
WINDSURF is still willing to dump code into the context, which gives them an edge in some cases (presumably at a cost of input tokens).
Windsurf is willing to spent to acquire customers (lower subscription cost, higher expenses for llm calls). Cursor has a huge customer base and is working on making it sustainable by a) reducing costs (see above) and b) increasing revenue (e.g. "Pro" requests for 0.05 with more input and output token).
In my experience, Windsurf was significantly more effective when working with a big codebase.
Windsurf goes looking into the codebase and learning context before attacking the problem in my experience. Often Cursor tries it's best to just guess the solution without context and only really resorts to going deeper when you tell it they fails.
I find if I tell Windsurf to look at something it will, Cursor I sometimes lay everything out for it and it just doesn't even read it.
Ultimately though once you run out of requests on Windsurf it's very weak without Claude though, and the top up requests are burnt through too quickly.
I trialled Cursor for a month and then Windsurf. Cursor read entire code files in while Windsurf would read the first 100 lines (or was it 50?), then the next 100, and often stop before it got to the part of the file with the method in which was needed.
So I went back to Cursor.
Have you checked Augmentcode.com? On reddit/youtube people are praising it for how well it handle large codebase compared to Cursor and Windsurf
Your other comments indicate you work there, you might consider mentioning that.
Boo I hate when people do that.
> Its the little things like having baked in instructions to install Windsurf on linux.
When I went to download Cursor the other day I noticed that they do not offer any .deb/.rpm packages and just offer the FlatPak (could be a Snap I'm not sure). This just tells me they really dont understand the community and just wanted to ship something for Linux and be done with it.
Windsurf has plugins for Jetbrains products, for vim, for emacs, for Visual Studio (not code), XCode, and even eclipse. They try to get as much of the market as possible, while Cursor focuses on the core functionality.
Whenever I tried Windsurf Editor, or their plugin for vim, and Intellij, it didn't feel polished at all. The basic function of autocomplete felt much much snappier on Cursor, and even on GitHub Copilot for vim/intellij.
There's a difference between understanding the community and prioritizing investments.
I'm sure Cursor has more than few devs that primarily use Linux...
> https://www.cursor.com/downloads
Linux builds are in the AppImage format.
Which makes a lot more sense to me than deb/rpm when it's just a single executable.
>just wanted to ship something for Linux and be done with it.
This what all developers want for a platform. They can release their software and not have to worry about some "maintainer" switching out dependencies out from under them introducing bugs and crashes in what they shipped.
Cursor ships as an AppImage.
appimage is more Linux than .deb/.rpm.
Of course it's a win, dude that cloned a GitHub repo is now personally a billionaire
Right wtf are we talking about. People are walking away with generational wealth.
$3B for a fork of an IDE which Microsoft keeps crippling by the day by making it's best extensions not work with forks (eg. C++, Python, C#, Remote SSH, etc)..
That's a oversimplified view. It doesn't matter if it's a fork. It has customers and paying ones. And it has a brand. That's more than enough. $3 billion would be peanuts for OpenAI
VSCode must have over 100 times the user base of Windsurf and Cursor combined. All Microsoft needs to do is implement a halfway decent version of the context management features these forks added. That alone would be enough to halt user migration.
For users who've already switched to the forks, the cost of switching back is essentially zero, especially if Microsoft begins introducing changes that break fork compatibility. In that case, the migration direction would reverse almost overnight.
> All Microsoft needs to do is implement a halfway decent version
and still MS could not build a chat App, they had to acquire Skype. Google could not build a social network.
If it acquired those customers in an environment where Microsoft was not enforcing their marketplace terms it very much does matter if they have a plan for supporting plugins in the future.
Are Cursor and Windsurf going to ask plugin devs to push to their own plugin stores in addition to VS Code's? Will they rally jointly behind a single open store? They need to have an answer to Microsoft here, and for the good of the ecosystem I hope they do have an answer, but customers will flee quickly if they lose access to all the proprietary plugins and to the broader ecosystem.
> Are Cursor and Windsurf going to ask plugin devs to push to their own plugin stores in addition to VS Code's?
They should and probably will soon, and if I were them I'd even consider giving plugin devs a cut of their paying customer subs if MS gets competitive about it.
> but customers will flee quickly if they lose access to all the proprietary plugins and to the broader ecosystem.
Agentic AI coding is more important to customers than VSCode's extension ecosystem. VSCode is who has to worry in this equation unless they ship the same tools in the next few months and heavily subsidize them.
VS Code is shipping agentic coding in the form of updates to GitHub Copilot. I haven't used it extensively yet since they added agent mode, but it's obvious that they're gunning for this market hard, and if I were into VS Code I would not personally choose to lose the ecosystem for marginally better agent mode.
The ecosystem will follow the users. If Cursor or Windsurf has better AI coding that’s where the users will be and the extensions will follow.
You’re in the minority if you favor manual coding + extensions over something doing your job for you.
But that's a false dichotomy, Cursor is far from the only capable agentic option. Personally I switched back to using VS Code with Cline + Github Copilot (just for autocomplete and included model access to Gemini Pro 2.5/Claude 3.5/7 that I can use with Cline).
given that they lose >$4B/year I guess everything is peanuts
OpenAI have $40 billion in funding from SoftBank for the next two years, so they can afford to buy Windsurf.
Is OpenAI worth the $260 billion valuation... No, of course not, they're losing >$4 billion a year.
That $40 billion is actively being spent being lit on fire to serve all the ChatGPT requests though. It's not just sat in the bank doing nothing.
> And it has a brand
Didn't they change names months ago? I know them as Codeium.
[dead]
It's easy to downplay as a fork because it's such a young product but ultimately if people use Cursor or Windsurf instead of VSCode then it is VSCode that needs to worry about being upstream from them and Cursor or Windsurf making their extensions no longer work with VSCode.
Microsoft "owns" OpenAI, which now owns Windsurf, which cloned VSCode.
I think it's going to be fine.
This is xAI buying Twitter, with extra steps.
We're reaching a point where we don't need to switch to another IDE (from VS Code/IntelliJ/insert-your-IDE-here) for "AI/vibe coding"
IDEs can support "AI coding agents" on their own.
The entire workflow for "AI coding agents" boils down to:
1. You write a prompt
2. The "agent" wraps it in a system prompt and sends it to the LLM
3. The LLM sends back a response
4. The agent performs specific actions based on that response (editing files, creating new ones, etc.)
Microsoft already started doing that with Copilot. And they have a vibrant ecosystem of VS Code extensions (I maintain one of them [1])
"AI agents" should be a feature, not a separate piece of software (IDE) that's integral to software devs.
[1] https://github.com/Kilo-Org/kilocode
So which of OpenAI's investors are also Windsurf investors?
does that mean that Windsurf will only support OpenAI models going forward? I doubt OpenAI will pay to have users use Gemini/Claude? Especially as all of these Ai coding tools (Windsurf, Augment, Cursor) are heavily subsidizing the users.
I wonder what Anthropic makes of this. Windsurf was like a top 3 customers of them, might be a big revenue blow too?
Maybe Anthropic will buy Cursor to level the field.
The queries made to other models is the juice of the data they will have access.
This feels absurd. When money clearly isn't an issue.
I think it's more of a time saver move by openAI - they can probably build something similar, cheaper – but, windsurf has established itself. Looking forward to see where this goes
Very strange move. They already have the models and their partner MS is providing the base editor.
So they’re paying 3bn for the integration and clients basically?
Seems pricey
they're paying to aquire places where you can sell tokens at a markup, because the future is multiple base models that are good enough for most user tasks where user gateways play the base model providers off each other and capture a lot of the value
probably a rare area I fully agree with HN on– the IP here seems weak and it's not hard to swap out code editors, nothing like tearing out Salesforce or other sales-driven tooling. and idk if first mover advantage actually means much in the next 10 years given how dynamic the underlying models are.
but undeniably these cos are all a great lesson in just how much cash lies in executing first/near first
As a Vimmer, I'm not into VS Code forks. I really like the goose CLI[1] though. Some untapped market potential right there.
1: https://GitHub.com/block/goose
all IDEs have vim plugins
Vim has IDE plugins. Terminal life forever
Recent announcements from OpenAI seem to indicate they know they're losing the race
You are referring to the nonprofit continuation?
They have certainly lost the monopoly.
They didn't even buy an IDE since windsurf is more like a VS code plugin.
So what was it exactly that was worth the 3B that they couldn't replicate themselves? Their prompts? Their training sets? Their users or user data?
Maybe time? OpenAI has access to basically infinite capital right now, if they believe this will be an importnat market and they could save a few months on launching this acquisition may be worth it for them.
I'd guess the prompts and employees.
I've found Windsurf more reliable/efficient than any other editors by leagues. How ever they have named the tools, crafted their prompts and generally how their internals reason is just on the money. I don't think that is easy to replicate, iterating on prompts over product releases whilst not pissing off your user base constantly is a feat in of itself.
Then perhaps it's about bringing in the human talent that wrote those prompts.
To be honest, Windsurf doesn't work like half of the time, so it's more likely their users, the data, and their branding/marketing potential.
Windsurf/Codeium plugin is at least 3 years old.
People seem to be pretty negative about this but of all the AI dev tools I've evaluated it's the only one that's felt meaningfully better than just using the web interfaces of the various frontier models.
I don't think it's good value for the money but pretending it's just a VSCode fork that wraps LLMs is underselling it. There's something they're doing that makes them better than Cursor, Claude Code, etc.
Would love to know what exactly Windsurf brings to the table that justifies $3B. Infra play? Specialized team? Or is this another move to consolidate talent before others can?
Valuation aside, Windsurf has built its own models [1] and boasts enviable enterprise distribution: $100M ARR, per TechCrunch [2]
[1] https://windsurf.com/blog/our-model-strategy [2] https://techcrunch.com/2025/04/22/why-openai-wanted-to-buy-c...
https://archive.ph/ocXFo
Pretty stupid move with Microsoft moving to put the kibosh on all of these proprietary vscode forks. Could be worth almost nothing in a matter of months...
why AI needs a text editor?
I don't get it.
With $3bn budget you can replicate it in few months, promote for free using your own stronger brand and you're left with roughly $3bn in the bank to do whatever you want.
Your sentiment is very common, which reminds me of those days where everyone claimed to be able to clone twitter/airbnb/dropbox/gmail/whatever over the weekend.
The real value lies in a successful execution, Windsurf is live and already has brand popularity with revenue stream. When it gets acquired by a titan, the die hard fans(vibe coders) will gain new trust that the product is not going anywhere and instead has solid future(we don’t yet know if oAI plans to do Msft style acquire, extend, kill yet). Cumulatively, this actually increases the value. Also Windsurf already has established enterprise revenue, hence brand trust and experience is already there.
In summary, an existing live and proven product is worth more any day over 100% uncertainty of building a sufficiently capable team to perfectly execute the same idea in specific deadline and also have the added burden of marketing, market penetration, user acquisition etc.
Windsurf is not present in enterprise.
Github copilot, vscode and apis through azure - basically everything through Microsoft - is.
Alternative to Microsoft's monopoly in enterprise that exists is open source.
Comparing this situation to twitter is more like if there was some chat api service, known more than twitter itself, that twitter is using/wrapping where other alternative clients exist, some with stronger popularity, some being open source.
> Windsurf is live and already has brand popularity with revenue stream
OpenAI has a much bigger brand.
Weren't they also in Cursor too?
This is probably a response to Claude Code, which is still experimental and terminal-only.
In my experience Claude Code is fantastic, both for answering questions about the codebase and coding.
OpenAI has Codex CLI https://github.com/openai/codex
I'd love it if windsurf were a single private Gist that is the prompt for OpenAI, and a load of people building UIs around it.
What's funny is the translation layer between your prompt <--> agent <--> LLM, that code was written by ChatGPT.
So OpenAI are paying for software which leverages other LLMs written by their own LLM.
We live in a topsy turvy world.
here is the thing, even those editors are relict of the pasts, the code is still in the center in these editors. thats something we need now, but not in the near (2 years, 5 years, 10 years?) future.
then the prompt is the coding, the reasoning is the execution, the code just an abstract layer that we do not care to much about i.e.: like assembly, machine instructions.
we know it exists, bit even here on hackernews i would guess only a small fraction know how it really works on a detailed level.
there will still be coding, instructions (prompt) -> execution (reasoning and AI code and code execution -> feedback (debugging to AI then and one point to the user)
bur actual looking at the code, well, thats only when this cycle annoyingly fails.
so current IDEs are still built from an code first mindset. this will not be the IDE of the future.
so basically OpenAI bought a Dinosaur
What is Windsurf's (or for that matter: Cursor, Cline, or CoPilot) moat? This seems like a great deal and timing for them.
From a customer point of view it makes sense to pay a fixed monthly price for both chat and coding, instead of having two separate subscriptions.
That makes a lot of sense, for the customers, but OpenAI is not profitable on even their $200 subscription. I doubt Windsurf is turning a profit either.
Buying a "bundle" should result in a lower price, as compared to buying both tools separately, making the loses worse. Unless they can reuse some of the same infrastructure and save a lot of money that way.
Question: has there been any announcements of bundling Windsurf with the ChatGPT $20/month package? (I could not access the linked article)
GitHub acquired for 7b, Windsurf a VScode fork + Agentic LLM… 3b$. I should be missing something.
> Windsurf a VScode fork + Agentic LLM… 3b$.
They don't have their own LLMs either, they've glued a 3rd party editor to 3rd party models. That's some expensive glue.
They have their own autocomplete model.
My bad, I was looking at the wrong thing. They use 3rd party models for chat but you're right, they rolled their own autocomplete model from scratch.
Feels like OpenAI is trying to double down on control and narrative at the same time
They seem almost exactly the same as Cursor, but even using the exact same rules, Cursor gives much better results than Windsurf (which performs below viable for me) - my test case was a complex Python project.
I like Windsurf for RSE, but it sometimes gets a little too excited which can take me out of the flow to undo stuff and get back into the groove of things.
The Claude integration is quite nice - I hope that doesn't take a step backward with the acquisition.
Here’s the crazy thing. All MSFT has to do is build in connections to every AI provider for VS Code and they win.
Embrace & Extend will never die.
> Embrace & Extend will never die
Spread too thin is often the final result of embrace and expand.
I don't get why people want the AI right in their editor. In another windows inside the editor, fine, but not inline with code I'm writing. It's super distracting to have AI auto complete pop up at random all the time. As always, typing speed, or speed at generating raw code, is not the bottleneck in programming. The crux remains design, in which case having the LLM on the side is just fine (if you use it for that).
There are some niceties about inline completion (like spelling out a log message that's obvious from the surrounding code) but I don't get the hype much beyond that.
Maybe I'm missing some feature though ...
If you have tried the completions in copilot, you are right. They are complete garbage.
Windsurfs on the other hand are much better. The only issue is that windsurf is super aggressive about them, but it is able to do do things like "the user made a change on this line, he most likely also want to make the change here".
Have you tried it recently?
AI autocomplete is the best thing I've experienced in developer experience in my career since git won over subversion.
I don't use LLM code prompting, but autocomplete is my jam. It's getting things right 90% of the time when I'm plumbing fields or refactoring. It makes life so much more pleasurable, and I say that as someone who is already using a statically typed language with robust IDE refactoring capabilities.
It's absolutely made me more productive.
I am happy with Copilot with VSCode..I do not think so, I would need to let AI generate the entire code. Even if I need, I copy/paste from Claude/GPT
Good for them, always rooting for startups who win.
That said, I have tried Windsurf multiple times, and it wasn't a pleasant experience compared to Cursor, which I've been using for more than 6 months as a paid customer.
M&A activity needs much more strongly regulated. Buying up potential competitors is how we get monstrosities like Microsoft and Alphabet.
In what world is Windsurf an OpenAI competitor?
A lot of this valuation must be aqui-hire and existing users, right? 6 months of development lead time can't be worth this much... can it?
I wonder how much of this is a data play for OpenAI as they work to improve language model performance on longer time horizons.
However crazy the 3 billion valuation is, windsurf's valuation is still very sane compared to that of Safe Super intelligence, who exists for less than a year, with no product, no roadmap,and virtually no hype, but is worth at 30 billion.
Big move by OpenAI—curious how Windsurf fits into the long game.
I use ChatGPT’s “work with” code helper and one of my biggest uses on ChatGPT. It’s a good first line before I pull out the big guns(APIs). Sadly the code canvas is rarely as it’s geared mostly for single page web app functionality useless demo tests. Maybe this is where Windsurf can come in
my summary here https://news.smol.ai/issues/25-05-05-cursor-openai-windsurf
Fortunately it is not the cursor. I am using the cursor and I don't want it to be sold.
Valuation lost its meaning in recent years :)
Bearish on IDEs after using Claude Code.
looks like VSCode
This may end up saving openAI. their models have no moat
The companies they are buying have even less moat than openAi
What’s the equivalent in the Vim world?
https://windsurf.com/vim_tutorial?extensionName=vim
Pure speculation without official voice.
I cannot pretend that I know what is going on - I don't.
I think the long-term play here is something to do with Agents and they are simply cornering the market because coding tools are part to it.
That being said, quick search around what people are building with these VIDEs reveals mostly landing pages that are actually not even that good. For the amount of money spent one could have easily bought a good template or pay someone to customise an existing one.
I don't know. Maybe I am dumb.
hmm..
Tell me what AI wrapper do I make that you would acquire my product?
Very surprising outcome, since OpenAI went after Cursor (twice) [0] And I originally thought that Cursor would be bought instead a day before the rumour [1].
It was smart for Windsurf to take the offer and to get greedy in this hype cycle. Unless Cursor is thinking that Anthropic or someone else will buy them for a lot more, its going to get extremely competitive as the switching cost for Cursor is zero and that ARR can disappear very quickly.
Copilot will attempt to destroy Cursor on price and functionality for however long they want to.
Very risky for Cursor at $9B valuation (which I think is overvalued and based on VC FOMO).
[0] https://techcrunch.com/2025/04/17/openai-pursued-cursor-make...
[1] https://news.ycombinator.com/item?id=43698819
sometimes products stick, like slack, dropbox, box cursor may survive
*not get greedy.
oh wow, meaning I won't need to pay for Windsurf? What do you think will be the monetization path for this?
Probably the other way around. Windsurf and co (Aider, Cursor) drives a heap of traffic to their API from which OpenAI actually profit. They just need to have their own tool to lock customers in their ecosystem.
the answer is always users and growth rate.
I need someone to convince me this isn't one of the biggest waste of money on an acquisition. If OpenAI can't build an official IDE for less than 3 Billion then what are they even doing? Windsurf can't have that high of a userbase that you feel the need to pay for it.
> one of the biggest waste of money on an acquisition.
I think that was when intel acquired McAfee for 8B in 2010.
MSFT buying Nokia for $7B is runner up. But at least it could have worked if MSFT hadn't burned it down. Intel and McAfee makes no sense at all.
We don’t know how OpenAI is paying. A lot of comments seem to be assuming this is an all-cash deal. We have no evidence for that.
These deals are mostly in stock, not cash. $3b cash is not something most companies can afford to part with, and additionally, making deals that are stock-heavy creates an incentive for the leadership of the acquired company to keep working towards the general interest of OAI, and not instantly retire.
> These deals are mostly in stock, not cash
How are you defining “these deals”? Most acquisitions of startup by larger companies in America over the last decade, at least, have been all cash.
I'd define them as a large company acquiring a private startup. Slack >> Salesforce was a cash-and-stock deal. Postmaters >> Uber, all stock.
In cases where the company being acquired is already publicly traded, those deals would have to be all cash as their shareholders would need to be bought out. IBM paid cash for Hashicorp, and Doordash will acquire Deliveroo in cash.
> I'd define them as a large company acquiring a private startup. Slack >> Salesforce was a cash-and-stock deal. Postmaters >> Uber, all stock
Okay, in that category of M&A in practically any category, the vast majority of deals are all cash. Deferred, for executives, in most cases. But cash.
> In cases where the company being acquired is already publicly traded, those deals would have to be all cash as their shareholders would need to be bought out
Not true. Preferable. Easier. Not not a requirement.
instinctively I agree but it's all about timing: if they try to build their own IDE and hired people it would probably still take a couple of years to get a decent product. I don't know about patents.
> if they try to build their own IDE and hired people
Oh, haven't you heard? Hiring people to write software is so last decade. Maybe they just didn't want to vibe code a Windsurf implementation and decided to buy a press cycle for $3B
you mean the company that spent $9B to make $4B in 2024? that openai?
i agree with you on this - it seems that openai hallucinates reality as much as their products do :-/
> If OpenAI can't build an official IDE for less than 3 Billion
It's funny money / made-up value. This is not $3B cash.
"$3B" should be in heavy quotes if this is paid in OpenAI shares.
probably paid in pro accounts
> 1,000 prompt credits/user/month
"hey Jim, can I use your credits? I have a deadline and I'm all out."
This is classic OpenAI - acquiring competitors rather than innovating internally. They're desperately trying to keep up with competition from Anthropic and Microsoft's GitHub, but throwing money at the problem is hardly a creative solution.
What's especially rich is the timing - right after OpenAI backpedaled on their restructuring plans due to "public pushback" (read: Sam Altman making yet another governance blunder). Now they're dumping billions into a tool that's essentially the same thing everyone else is building.
Ok, now I have a question: Will OpenAI keep Windsurf open to third-party models, or will they limit it to their own models only?
damn.
openai just seems to have a hole in their hand they keep temporarily patching up with new investor money
Now their models may have limits on how VS code and Cursor use it. Competition heating up!
Who are these people that give OpenAI all this money? Aren't Microsoft, SoftBank, Nvidia publicly traded? Don't they owe a fiduciary duty to their investors? I'm surprised it's legal to just hand over a blank check to random private companies to make nonsense purchases. This isn't going to end well.
If I were any of these companies I would be suing OpenAI to try to get my money back. Thrive, ARK, Tiger and the others can pound sand
Long term investments exist.
An _ide_ sold for $3B? VCs and other early investors got their 1000% ROI on this one.
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
[dead]
[flagged]
[flagged]