Comment by theahura
14 hours ago
Lots of people in the comments talking about how this is about training data, but surely this is actually about hiring competent people after the mass exodus/firing at xAI?
14 hours ago
Lots of people in the comments talking about how this is about training data, but surely this is actually about hiring competent people after the mass exodus/firing at xAI?
Whoever thinks the talent pool is this limited that it requires offering Cursor of all places $60B is pattern-matching so hard they might as well be a quilt.
It's not about the talent pool at all.
The AI bubble music stops when one of these former-darling companies has a complete crashout rather than a "successful exit". The investors keep investing because even largely-failed products get acquihire paydays
hi, im the quilt.
Note that Meta paid ~16b for Alexandr Wang, and Google paid ~3b for the windsurf executive team. You are making a category error -- the talent pool isn't "ML researcher" it's "competent leader"
These guys aren't 100000x leaders, they're investment vehicles. They're nfts in human form.
No, it's "I'm a more important person if I ok deals with big numbers" that always happens in a bubble.
They could offer $20 million dollar signing bonuses to every Cursor employee if they wanted to hire them away and it would be much cheaper.
They’re buying the customers and the brand.
Buying the customers seems though, when it looks like they migrate to whomever offers the steepest subsidies.
A brand he’ll promptly burn to the ground by renaming it some garbage with an X in it.
That's quite a pricey acquihire
$60 billion worth of competent people?
Are cursor developers “competent” in creating frontier models? Aren’t they just using other company’s models?
I think composer has currently by far the best price to performance ratio for coding (not counting subsidized subscription cost by OpenAI and Anthropic). It's based on Kimi K2, but I think it's fair to say, that their RL really sets it apart from the other open weight models.
Training any large model at scale is hard, and Cursor has trained several including agentic ones. https://cursor.com/blog/composer
60b?
10b active it seems