@adamaoc @aakashgupta Because if you are transparent you'd word your announcement 'We've fine tuned Kimi K2.5 to be much better at coding' instead.
English
Cole Phelps
423 posts
Cole Phelps
@XmarksTheDoubt
Running background checks on hype and headlines.🕵️ Based commentary on all things tech, AI and science. Here on ⓧ to doubt. 🎮
Присоединился Nisan 2023
92 Подписки40 Подписчики
@XmarksTheDoubt @aakashgupta I never understood why this was such a big sticking point for people. Cursor specifically has the right kind of data to take any model and make it better at coding. If you look at benchmarks from K2.5 vs Comp2, there's a huge difference.
English
This is wild. SpaceX now has the right to BUY Cursor for $60B.
Or pay them $10 billion to walk away. To put it in perspective, Cursor was worth $9.9 billion total in May of last year.
Let's have a closer look at the numbers.
Start with the $60 billion. Cursor was already raising money this week at a $52 billion valuation from a16z and Nvidia. The Elon offer sits 15% above a number that was already on the table. The next round priced in, with a one-year fuse.
The $10 billion is the real number. That's what SpaceX pays even if it walks away and never buys the company. The walk-away fee alone is more than the entire company was worth 12 months ago.
Now the strategic logic. Cursor stopped being just an editor in March. They shipped Composer 2, their own model, and it beat Claude Opus 4.6 on Terminal-Bench at one-tenth the price. The catch is that frontier coding models need frontier compute, and the only labs with frontier compute are the same ones building competing coding products. OpenAI shipped Codex. Anthropic shipped Claude Code. Google has Gemini CLI. Cursor was renting capacity from every company trying to kill it.
Colossus is the way out. 230,000 GPUs in Memphis today, 1 million by year end, the biggest training cluster on Earth. The Information already reported Cursor is renting tens of thousands of those chips to train Composer 3. SpaceX is also building Grok Code, so they're not a clean partner. But xAI losing the coding race to Cursor is a better outcome for SpaceX than Cursor losing the coding race to OpenAI.
The trade Cursor made: gave up the right to be acquired by anyone else for one year. Got training compute at a scale no other lab would sell them. Got $10 billion guaranteed if Elon walks. OpenAI tried to buy Cursor in early 2025 and got rejected. Cursor stays independent for at least 12 more months and gets to train on the biggest cluster on earth doing it.
Elon just bought a one-year call option on Cursor for $10 billion.
That's the deal.
English
@realantonmaier @LimitingThe Indeed, within 24 hours of release 'their own model' turned out to be fine tuned Kimi K2.5, which this and the original post omits.
English
@LimitingThe Wasn't there news that cursor rebranded other LLM providers Code?
English
"Cursor stopped being just an editor in March. They shipped Composer 2, their own model, and it beat Claude Opus 4.6 on Terminal-Bench at one-tenth the price. "
However, cursor didn't have enough access to compute, which is why they're partnering with SpaceXAI.
Worth a read.
Aakash Gupta@aakashgupta
This is wild. SpaceX now has the right to BUY Cursor for $60B. Or pay them $10 billion to walk away. To put it in perspective, Cursor was worth $9.9 billion total in May of last year. Let's have a closer look at the numbers. Start with the $60 billion. Cursor was already raising money this week at a $52 billion valuation from a16z and Nvidia. The Elon offer sits 15% above a number that was already on the table. The next round priced in, with a one-year fuse. The $10 billion is the real number. That's what SpaceX pays even if it walks away and never buys the company. The walk-away fee alone is more than the entire company was worth 12 months ago. Now the strategic logic. Cursor stopped being just an editor in March. They shipped Composer 2, their own model, and it beat Claude Opus 4.6 on Terminal-Bench at one-tenth the price. The catch is that frontier coding models need frontier compute, and the only labs with frontier compute are the same ones building competing coding products. OpenAI shipped Codex. Anthropic shipped Claude Code. Google has Gemini CLI. Cursor was renting capacity from every company trying to kill it. Colossus is the way out. 230,000 GPUs in Memphis today, 1 million by year end, the biggest training cluster on Earth. The Information already reported Cursor is renting tens of thousands of those chips to train Composer 3. SpaceX is also building Grok Code, so they're not a clean partner. But xAI losing the coding race to Cursor is a better outcome for SpaceX than Cursor losing the coding race to OpenAI. The trade Cursor made: gave up the right to be acquired by anyone else for one year. Got training compute at a scale no other lab would sell them. Got $10 billion guaranteed if Elon walks. OpenAI tried to buy Cursor in early 2025 and got rejected. Cursor stays independent for at least 12 more months and gets to train on the biggest cluster on earth doing it. Elon just bought a one-year call option on Cursor for $10 billion. That's the deal.
English
@0xrwu @LimitingThe Definitely talking out of your ass. No researcher at Cursor is any better than any that are left at xAI.
English
The structure of the deal is pretty interesting here. I think what’s happening is:
1. xAI is having trouble training a SOTA coding model (hence cofounder departures), bunch of idle GPUs
2. Cursor doesn’t have capital to blow on a $5B training run to compete with Codex/Claude
3. xAI says to Cursor: use all the GPUs you want at cost and get to a SOTA coding model, as long as we have an option to buy you
4. Cursor also gets a free option: train a model better than Opus and get bought out for $60B, or get $10B that pays for all the GPUs you rented
Win win
SpaceX@SpaceX
SpaceXAI and @cursor_ai are now working closely together to create the world’s best coding and knowledge work AI. The combination of Cursor’s leading product and distribution to expert software engineers with SpaceX’s million H100 equivalent Colossus training supercomputer will allow us to build the world’s most useful models. Cursor has also given SpaceX the right to acquire Cursor later this year for $60 billion or pay $10 billion for our work together.
English
@ericweinstein Didn't Terrence Tao release a recording of him working with Claude?
English
The off-the-shelf product of Top-Tier Commercial AI does NOT work for mathematical/theoretical physics.
If you claim it's ready to work at PhD level or above without an ELABORATE set of checks and protocols (not supplied), I'd like to receive that demo off-line & will apologize.
English
@SebastienBubeck @sama 'What I believe to be true' is a very specific and clever use of words when it comes down to admitting you were wrong in the future
English
Just to say it out loud, @sama is an incredible leader, and just a ton of fun to work with. I have never seen such a gap between the public image and what I believe to be true (arguably I don't know that many people with Sam's level of public profile 😅)
Sam Altman@sama
I wrote this early this morning and I wasn't sure if I would actually publish it, but here it is: blog.samaltman.com/2279512
English
What happened to Sam Altman and his family is really awful.
It is hard to reconcile his call to “de-escalate the rhetoric and tactics” with his implication that a piece of critical journalism (Ronan Farrow and Andrew Marantz’s New Yorker article, presumably) was responsible for this.
Sam Altman@sama
I wrote this early this morning and I wasn't sure if I would actually publish it, but here it is: blog.samaltman.com/2279512
English
@bryan_johnson Is there a specific position that promotes your longevity?
English
@fandompulse That's because shallow work sells better and he wants more 🤑
English
Project Hail Mary writer Andy Weir on social commentary in books:
"I dislike social commentary. Like… I really hate it. When I’m reading a book, I just want to be entertained, not preached at by the author. Plus, it ruins the wonder of the story if I know the author has a political or social axe to grind. I no longer speculate about all possible outcomes of the story because I know for a fact that the universe of that book will conspire to ensure that the author’s political agenda is validated. I hate that."
"I put no politics or social commentary into my stories at all. Anyone who thinks they see something like that is reading it in on their own. I have no point to make, and I’m not trying to affect the reader’s opinion on anything. My sole job is to entertain, and I stick to that."
"To that end, I also don’t talk about my personal political opinions publicly. I don’t want readers to even know, honestly. I don’t want that in the back of their minds as they read my stuff."
Is this why he has the #1 sci-fi movie in decades?
English
@TheAmolAvasare @AnthropicAI @grok what's would you estimate as a subscription / API costs for releasing Mythos to public? Could this be unfeasable economically and safety is just a nice public narrative?
English
I run growth at @AnthropicAI. My job is to get our models into as many hands as possible.
Mythos is by far the strongest model we've ever trained, and it's been a step-function change for us internally.
It hit 93.9% SWE-bench Verified, found a ton of critical bugs including a 27-year-old one in OpenBSD, yet we chose not to release it broadly.
Project Glasswing is what "safety as a product decision" actually looks like. Proud of this one.
anthropic.com/glasswing
English
can a fucking notification bell update correctly when I click it off or is it too much to ask of the elite engineering team working on the app?
English
@DiscussingFilm I think you made a very unfortunate casting choice for a character that includes Ape in their name
English
First trailer for the ‘HARRY POTTER’ series.
Releasing this Christmas on HBO.
English
Went from scrolling 24/7 to not even wanting to log in. X just hasn’t been hitting the same lately.
English
English
True.
New Roadster unveil probably in late April.
Massimo@Rainmaker1973
Today in 2008, Tesla began the regular production on its first car, the Tesla Roadster. Also, a reminder that there's one Tesla Roadster in an elliptical heliocentric orbit crossing the orbit of Mars.
English
Watching Nvidia destroy the DLSS brand in real time is quite something.
I wonder if they were expecting the "AI slop" reaction internally, or whether they had fully drunk the Kool-Aid
English
If you want me to keep my account open you may have to actively protect me against harm, because the design of the platform does not. (But I am ok with keeping the account protected, shutting it down or taking it elsewhere.)
English
@PatrickHeizer What's expensive and how much exactly? Sequencing a genome is cheap and the regular cancer treatments aren't pennies either
English
Sorry to be the downer because this is an impressive story in some senses.
But it is ~trivially easy to make a single mRNA vaccine. It's not hard.
I cure mice of various cancers with various therapeutics all the time. I've made mice lose more weight in a month than tirzepatide does in a year.
What is hard and expensive is proving its BOTH safe AND effective **in a randomized and controlled study in humans** while ALSO manufacturing it at clinical scale and grade.
I am happy for this man and his dog. It is impressive.
But y'all are overhyping it.
Séb Krier@sebkrier
This is wild. theaustralian.com.au/business/techn…
English
@demishassabis My dog suffers from meningioma and probably has weeks to live at this point, I guess this procedure isn't possible without tumor biopsy?
English
@peterwildeford xAI will catch up this year and then exceed them all by such a long distance in 3 years that you will need the James Webb telescope to see who is in second place
English
Based on the data I see, I think:
- Anthropic🇺🇸/Google🇺🇸/OpenAI🇺🇸 all ~tied
- Meta🇺🇸 / xAI🇺🇸 each ~7mo behind
- Moonshot🇨🇳/- Deepseek🇨🇳 / zAI 🇨🇳 / Alibaba🇨🇳each ~9mo behind
- Mistral🇫🇷 ~1.5 years behind
- No other companies competitive
Ethan Mollick@emollick
Both xAI and Meta seem to be falling behind, based on the Grok 4.2 benchmarks and this reporting. Frontier AI models are really a three way race at this point.
English