> We’ve grown to a team of over 300 engineers, researchers, designers, and operators
That last word, operators, I have seen used multiple times over the past couple of weeks to refer to managers and politicians. Is that the usage here too? If so, is this a new trend in the tech world? I’ve certainly heard of “political operators” in TV shows about Washington DC, but the usage in tech is new to me.
Prob burning through 200% of revenue which I've seen elsewhere. But they also probably spend a fair amount training their own model. I don't think it's foundation model. But it's pretty fair to assume that $1bn revenue is about $2bn to Anthropic/GPT/Grok
I barely use the autocomplete features of Cursor, and for agentic coding, Claude Code blows Cursor Agent out of the water. I don't think Cursor has anything that cannot be replicated in a week or two other than the first mover advantage; certainly not an advantage that cannot be justified at 30B+ valuation.
Comments like this remind me how much of the ai/agentic ecosystem is based on people's personal vibes and emotions
I've seen very little, meaningful difference. They all have their quirks, things their good at / bad at. The underlying models are very similar as well
I think the problem is that it can be a full time job on itself to try to test all of the available alternative. The models and editors, cli tools that aims for "increasing developer productivity using LLMs" comes and goes much faster than most people can even track.
I think what you are say is true too but another angle is that people use these tools in different way so they yield different results. Hell even the expectations are different. Someone prompting for some React components will much happier with Claude sonnet 4.5 than me. I do heavy GPU programming and scientific computing stuff where LLM will mostly give you hallucinating answers 80% of time.
I like Claude Code in the terminal. For me it's so good it don't need IDE integration. I'm just using emacs and magit to navigate the code out of band.
I have both Cursor and VS code copilot in my work machine, but haven't really felt the need to use Cursor. VS code agent mode with Claude Sonnet is actually taking care of everything so far, plus I get to keep using my old launch config and debugging workflow.
> We’ve grown to a team of over 300 engineers, researchers, designers, and operators
That last word, operators, I have seen used multiple times over the past couple of weeks to refer to managers and politicians. Is that the usage here too? If so, is this a new trend in the tech world? I’ve certainly heard of “political operators” in TV shows about Washington DC, but the usage in tech is new to me.
> Today, we’re pleased to announce a new round of financing: our Series D of $2.3B at a $29.3B post-money valuation.
> We’ve also crossed $1B in annualized revenue
A 30x revenue multiple on (presumably) relatively low-margin revenue is certainly punchy.
One wonders how much of their $1bn of ARR they're paying straight through to Claude/Anthropic.
300 employees, let's say average salary of 300k?
$90m in employee expenses so that's neglible.
Prob burning through 200% of revenue which I've seen elsewhere. But they also probably spend a fair amount training their own model. I don't think it's foundation model. But it's pretty fair to assume that $1bn revenue is about $2bn to Anthropic/GPT/Grok
They haven't built their own editor, they haven't built their own models; what have they actually built?
Well, they delivered something that is usable and useful for me and my team, and a lot of people I know, and I guess that’s what counts in business?
I barely use the autocomplete features of Cursor, and for agentic coding, Claude Code blows Cursor Agent out of the water. I don't think Cursor has anything that cannot be replicated in a week or two other than the first mover advantage; certainly not an advantage that cannot be justified at 30B+ valuation.
Have you tried Zed? Cursor is terribly slow and buggy.
I mean, they have built their own model: https://cursor.com/blog/composer
And presumably they'll use the funding to build more than just a modified VSCode.
most likely a finetune of existing model
What they the money for?
The average series D is 50-100M. This is 2.3B.
I'm wondering if AI coding companies almost NEED to be this capital heavy to pay for the massive LLM costs.
They propably burn something in the order of 50M-100M per month in LLM API costs for models like Sonnet 4.5. So the answer would be: Yes.
I love Cursor.
I’m greedy to ask but is there a better alternative? Hard for me to imagine. I tried Copilot was no where near as good.
Comments like this remind me how much of the ai/agentic ecosystem is based on people's personal vibes and emotions
I've seen very little, meaningful difference. They all have their quirks, things their good at / bad at. The underlying models are very similar as well
I think the problem is that it can be a full time job on itself to try to test all of the available alternative. The models and editors, cli tools that aims for "increasing developer productivity using LLMs" comes and goes much faster than most people can even track.
I think what you are say is true too but another angle is that people use these tools in different way so they yield different results. Hell even the expectations are different. Someone prompting for some React components will much happier with Claude sonnet 4.5 than me. I do heavy GPU programming and scientific computing stuff where LLM will mostly give you hallucinating answers 80% of time.
full agreement, the "my choice is the best" discourse has become quite tiring, I feel the same way about the rust stan'n
I like Claude Code in the terminal. For me it's so good it don't need IDE integration. I'm just using emacs and magit to navigate the code out of band.
I have both Cursor and VS code copilot in my work machine, but haven't really felt the need to use Cursor. VS code agent mode with Claude Sonnet is actually taking care of everything so far, plus I get to keep using my old launch config and debugging workflow.
Cursor has the best tab-complete.
For agentic coding, people tend to prefer Codex or Claude Code, but I haven't heard many opinions about Cursor's new Composer yet.
Of course, Sourcegraph is better.
What does Amp do better than Claude Code or Codex? I find the concept pretty appealing but the pricing is a bit scary.
It has better search and context management under the hood, which matters for big companies.
But you can also see quantitatively that Sourcegraph produces the most accepted code: https://www.theinformation.com/articles/openai-catching-anth...
Crazy for a 2.5 year old company
Nothing to see here. Just your average 2.5 year old start-up worth nearly as much as Ford or VW.