And they’re about to buy Tesla for an ungodly sum, sweeping another failure under the rug, this time with every index fund holder bailing them out.
This is just hog wash.
They are buying it for 60bn worth of SPCX Class A stock. This is Cursor’s board taking anything they can get for a company that has no future. Not a criticism against Cursor, but their market is standardising on Claude Code, and their value would rapidly head towards zero.
Class A stock has no voting rights. This is the investors of Cursor literally saying “being on Musk’s train to the moon is literally better than whatever future we have right now”.
The AI funding model.
True but they’re still making a bank and while the best exist was months ago the next best exit is now. I’d very happily take 60b worth of meme stock for a product with no future.
So the companies just get random values these days? It used to be that you actually had to generate some sort of profit before you were considered a valuable company.
I have my own business, so I just declared myself a billionaire. Now the government can send me billions in government contracts.
Is USD just Monopoly money now?
The last guardrail the US had was Jerome Powell.
You can use US$ to Warsh your ass from now on.
Always has been.
Wilson brought back the US Central Bank (aka the Fed). Andrew Jackson killed the bank to free us from their clutches.
Nixon killed the gold standard so we could be the worlds central currency printing press and export inflation to the whole world and bully everyone into submission and extract their resources.
Wall st finished the financialization and commoditization of everything on the world so they could own it and gamble with the future while everyone else pays the bill.
👨🚀🔫👨🚀
The US economy is just cooked, none of it is real or makes any sense.
Hello, 1929 crash out upcoming.
Nazis buying up more shit. Fuck all these clowns. Free Luigi.
Why can’t he download cursor for free? What an idiot.
My cursor came free with my operating system, although I did have to buy my own mouse.
$60 billion.
Minecraft was acquired for 2.5 billion in 2014.
I kinda remember reading how the purchase of Macromedia by Adobe was like the biggest acquisition ever at 6 billion or so back in the late 2000s.
atleast minecraft actually worth that much, probably
It’s actually worth wayyy more. Microsoft already broke even just 2 years later (2016). It’s estimated they made 8.5B total since acquisition.
Cursor will likely never make 8.5 billion let alone 60.
But Microsoft didn’t plan on that. They only bought Mojang because they needed to launder offshore money from Europe.
In case you guys haven’t figured it out, Musk is just a tool of Wall Street, whether these companies make money or not is completely irrelevant.
Okay, that put things into scale for me.
Almost like someone is lying about the inflation numbers for the past decade
Because all space companies need a social network and an ai coding company. It just makes sense. And your pension needs to buy this very logical company because it’s large enough to be in the index.
Yeah, SpaceX is done in terms of exploring space musk is just cashing it.
SpaceX is 80% an AI company. More now, after acquiring Cursor. The space stuff is only the reception desk.
deleted by creator
Isn’t the only part of company which is profitable space stuff?
SpaceX is 100% a bullshit company. FTFY
This is the sort of comment where I wish Lemmy had the option to block everyone who upvoted it.
edit and I wish the ones who downvote this also blocked me
“We do social media, rockets, and AI” Yes this company is a very logical well thought out entity who’s losing $3 for every $1 we make. We are growing only at 33% a year, yet are valued at 25x other companies with better growth and actual strategies . If you believe this shit I’ve got a bridge to sell you.
The original Founders had a lot of good ideas and did a lot of hard work. Unfortunately they let a clown come in and take them over and turn everything into a circus. Where people can say that sort of thing 100% accurately now. Tesla and SpaceX are both punch lines realistically anymore as is Twitter.
You realize Elon is the founder of SpaceX.
The lead who designed the merlin engine turned down being a founder as he wanted a stable income vs the low pay and risk of working for equity/founder status, so he was an employee. (Edit and I don’t blame the guy for that choice either, odds of a private rocket company succeeding was incredibly low)
SpaceX is not a punchline.
SpaceX is an infrastructure company. Deeply dependent on government good will and contracts. It leases it’s data center compute to actual AI companies.
They aren’t deeply dependent on government contracts anymore. Starlink will be able to support them. Government good will, id say yes.
Starship is currently a money pit on the launch side, and will make or break that the company depending on if it works.
Its leasing excess capacity, but that might not always be the case, but IMO that business model is probably better than the selling an LLM model one, but if you run out of people willing to lease it, youre still left with a pile of hardware you gotta do something with.
If they stop leasing, xAI becomes a money pit too.
I agree. But suffice to say if NASA walks away from SpaceX because they fail to deliver, it’ll make an outsized dent to their perceived value.
For sure, if something happened and NASA walks away, or reduces their interactions with SpaceX that will have ripple effects beyond what they currently provide.
This truly is the weirdest timeline. The only division of this shit company that brings in any money is the space and rockets division. However, the technology divisions are a money pit.
brings in any money is the space and rockets division
Not even that. Only StarLink brings in profit.
I’ve come to call it SpaceXitterAI
Perhaps I’m missing something, but isn’t Cursor simply a VSCode fork with some AI extensions integrated into it? Hate aside, what is the actual technical value of the product?
They don’t care at all about the vs code fork (probably his colossus data center with 89% vacancy can vibe code a clone in a few hours)
They care about the subscribed users and how they can enshittify their experience by removing access to Claude opus and introducing grok the Nazi vibe coder
Early on in the AI game, before Vibe Coding was even a term, Cursor was the only tool that could connect VsCode to an AI model. Since then, there are now dozens of tools to do this for you. Cursor basically has no unique features, the only value it has as a company is an existing userbase and an established dev team. The worst part is that they are built on top of VSCode, which means their existence hinges on Microsoft benevolently sharing one of their most valuable products, and Microsoft has made several moves to try to kill the Cursor userbase and bring them back to VSCode.
Cursor is one of the worst investments you could make right now. It’s surrounded on all sides by rivals trying to eat their lunch, and they have basically zero IP to protect their market advantage.
Even back then, copilot already existed if you didn’t mind choosing one of their models. Cursor was at most 2 months ahead in features, but already charging twice as much as copilot.
He is buying the users that cursor had. But these users can switch to another AI harness any day they wanted to.
While it is nowhere near the value Elon paid. It is a tool for easy coding with ai. Like the others thousands that out there. I mean spacex buying companies left and right at prices that make no sense is weird. Spacex being valued that much when all they do is burn money at a rate of several states is confusing.
Elon getting richer by the second with all that nonsense is suspicious…
Maybe he’s buying with overvalued useless SpaceX stock. SMH
Ding ding
It’s probably gonna be fine, Elon is just doing what Grok tells him to at this point
I mean even grok says that he is trying to manipulate the stock market.
/s*
Haven’t used cursor but I assume it’s like other harnesses where the special sauce is in how it manages context, schedules sub agents, feeds them context, enforces standards, uses tools and skills etc. that make it better then just directly prompting the model.
For example you prompt opus directly with “refactor the auth flow” and it’s going to try and “one shot” it and produce the code from that prompt. Whereas a harness has instructions to say
- Research the current implementation
- Search the web for standards
- Ask the user questions on how they want to do it …
Which produces way better results
Claude code’s harness isn’t the best, but it’s still better than just promoting the model directly. One thing cursor has going for it is the vector embeddings of your codebase to make the harness more token efficient, but I still switched from cursor to managing parallel Claudes in cmux a few months ago and have had little reason to go back.
They have their own model called Composer that’s specialized for coding.
Personally, I think generic Opus 4.8 beats it, but it is quite cheap and fast. I used Composer via Cursor for a few months in late 2025. I hear they have improved the model somewhat since.
It’s fine-tuned Kimi 2.5
They didn’t train this up for themselves.
It’s an OK model that is fast and cheap. It outperforms other models of the same price class, but it’s nowhere near Frontier model capability.
They are also claiming to be training their own new model on SpaceX’s infrastructure. I don’t particularly trust them, but I’ll give the citation anyway just for additional info.
They entered into an agreement to buy Cursor, with an expected time for the merger in Q3. It’s an all stock transaction, the number of shares that will depend on the SpaceX share price, to make up the 60 billion USD. Also they will pay in Class A shares (so the shitty ones with nearly no voting power)
“Cursor, help me figure out why this request is failing with a CORS error.”
Thinking…
Reviewing sources…
- “Elon Musk on X”
- “Elon Musk Twitter acquisition” townhall recording
- “Programming best practices” and “Elon Musk”
Done.
It looks like everything needs to be thrown out and redone. The whole stack. It’s not good, it’s bloated, it was made by overpaid salary workers.
“What’s wrong with the stack?”
The whole stack. Just the stack. All of it.
Sure would like some food, affordable housing, and healthcare…
We are moving away from cursor. Got too expensive
Interesting, where are you moving to?
We are looking into open router and kimi 2.7, mimo, glm 5.2 models.
You can still use opus/sonnet models all the same, they are just wicked expensive.There are tools like opencode and kilocode for coding.
Have you tried composer-2.5? Supposedly it’s fast, decent, and far cheaper than sonnet.
It’s based on kimi 2.5. but since then kimi 2.6 and 2.7 came out already. Yeah I’ve used it
I had to move to Kimi K2.6 because Claude models are too expensive to use. BYK from OpenRouter. I have not tested GLM models.
As they stop subsidizing, everything is going to get more and more expensive until they price everyone out. If they are actually “selling shovels for a gold rush,” they are cooked. A lot of companies can’t afford to use Ai heavily anymore. Customers are going to dry up.
Exactly. It’s hilarious just couple days back our ceo pulled 4 devs out and congratulated us on spending 2k a month each and apparently blowing company limit, I’m now tasked with finding alternatives hahaha. Definitely openrouter and kimi 2.7 and glm 5.2 are on my roadmap. I will need to do a presentation for ceo
I am surprised that anyone who can make decisions realized the June Ai bill was going to be crazy. I was betting it would take until July, when the crazy June bill is charged, for companies to realize they can’t afford frontier models.
Local coding LLMs are going to be the hot new commodity this year.
Have you looked at hardware prices?
Harnesses for them is the key for me. Check out llama conductor
https://codeberg.org/BobbyLLM/llama-conductor
Smaller models being used more efficiently.
Already is, take a look at devstral, qwen3.6, deepseek coder. All can be run on a hugh end GPU and if you’re a developer you likely have one.
The vast majority of users ain’t running anything but 27b max, more likely 14b, and that shit just ain’t nearly as good as older saas models much less dominant like opus. Maybe for small shit but complex talks just ain’t fitting on home hardware.
Completely agree, I forgot to mention that part. I am testing a few models ranging from 18b to 26b on my 7900xt. It is far from “make this complete system”, but it can handle some smaller tasks. I think that will be the end goal anyway since cloud models fail a lot at maintainability, security, and other higher levels of thought that goes into coding. They can make a convincing prototype but I wouldn’t hook it up to production.
Local models are already functioning well as a force multiplier. It can help explain logic, do minor refactoring, debugging etc. but with a bit of latency. I do think this is where we’re headed since the frontier models required for generating a full prototype can’t make production quality code and it is prohibitively expensive to do so. As far as I’ve heard, they’re generally running spending ten times as much as they earn per token.
My guess is the next big thing to come out is, we can probably squeeze a lot more reliability out of smaller models. But their workflows, context, validations, etc will need to be very tightly optimized.
I can see harnesses coming with their own highly specialized lightweight models in the future. Some for very efficiently converting a basic prompt into chain-of-thought steps. Some for very efficiently determining relevant parts of a repository. Some for… a lot of highly specialized stuff. Then the harness would orchestrate these under the hood, reducing the cognitive load placed on any larger generalized LLMs. Those “larger generalized LLMs” could be something like 12b parameters.
Hopefully, soon after, we can start benchmarking how much different harnesses and augmentations improve baseline model performance. Ideally, in the long run, with a deeper understanding of how to tailor harness to workload and produce more procedural determinism. Then we can start configuring harnesses like data pipelines and run them through higher-level orchestration like Airflow too.
Sadly, that’s true
Tried to refactor a spaghetti code state machine and thought, well, AI should handle this well. All the logic is there, just separate it into small functions to clean up the large one.
None was able to, alone because of the context window already
To be fair though, I tried Mistral online and it also stumbled around. ChatGPT was a complete clusterfuck - haven’t tried Claude.
To be even fairer… it’s a really large state machine, which was written on site during a fever and in stress - so… To defend myself a bit as well, how it even came to that ;-)
But seems, I’ll need to go through this myself
Actually thought, that this would be a perfect example for using AI…Yeah LLM’s can help with many tasks but then there are times they just spout nonsense, or syntactically correct nonsense, the model size and context window just changes when they hit their limit.
Sometimes you have to call it quits, and try another way.
Why would a developer likely have a high end GPU? Writing code doesn’t use a GPU.
There is a significant overlap between developers and gamers.
Most developers use their work provided machines, which aren’t gaming machines with giant GPUs because again, GPUs don’t help development at all.
Also developers often want more ram, and if youre on the mac side, the M series ram works as video ram for loading and running models, so there’s a good chance you can already run something better than is typical of others, and apple is focusing on this by adding more NPUs and increasing memory bandwidth. They arent good at training, but can do inference.
I’m on a MacBook with M2, 32GB ram. Literally just tried:
- gemma4:12b - very slow, unworkable
- qwen3:8b - very slow, unworkable
- qwen2.5-coder:7b - slow but workable. Doesn’t use tools properly in OpenCode.
Well, I guess I’ll try again next year.
For context: my home pc is running gemma4:31b just fine. It’s also a beefy ass desktop, though.
Are you running an mlx model? If not, try that. My m4 macbook runs qwen3.6-35b-a3b lightning fast. Has its issues, but fast nonetheless.
What kind of context length can you get with that, and how much ram?
You might be doing something wrong, models that size shouldn’t be that slow if properly configured on a 32gb m2
You need a metal optimized client and model, not the same models you’d run on your desktop machine.
