Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Cursor Composer 2 is just Kimi K2.5 with RL

twitter.com

Post reply on HN
Print view View original

141–150 of 180 posts

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#141

Post by varispeed »

merlindru wrote: post #124
varispeed wrote: post #122

I noticed something strange with Cursor lately. When I am using Opus 4.6, sometimes it is giving ridiculously dumb answers as if they were actually using something like Qwen with a prompt to present itself as Opus. I have to close the session and start again hoping I'll get actual Opus.

There's no shot they're doing that. Would be suicide as soon as anyone notices, and by the looks of it, they didn't even clean up the URL here to "hide" the fact that this is Kimi K2.5 so i doubt there's any grand conspiracy here. What's way more likely is that Opus has been quantized by anthropic or something similar. Or that Opus was updated and didn't work well with Cursor's harness after. Or a token caching issue…

Well, I noticed. Though I am too busy to make a fuss about it.

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#142

Post by simonw »

leerob wrote: post #138
simonw wrote: post #63

I'm annoyed that we still don't know for certain which base model they used for Cursor 1. This feels really rude to me. I have no problem with them fine-tuning open weight models to create their own - they are getting great results, and Cursor's research term should be respected for that. But deliberately hiding the base model they use is disrespectful of the researchers who created that model.

We used a Kimi base, with midtraining and RL on top. Going forward, we'll include the base used in our blog posts, that was a miss. Also, the license is through Fireworks: https://x.com/Kimi_Moonshot/status/2035074972943831491

Thanks for sharing this, I saw Cursor are committed to publishing this now which is great to see.

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#143

Post by geocar »

hakunin wrote: post #118
miroljub wrote: post #72

Earlier quoted context omitted.

> Cursor Composer 1 was Qwen and this is Kimi. IDE is based on VSCode. The entire company is build on packaging open source and reselling it. The question is, where's the outrage? Why are there no headlines "USA steals Chinese tech?" "All USA can do is make a cheap copy of Chinese SOTA models". > So funny to see Twitter go wild saying "a 50 person team just beat Anthropic" blah blah. Well, if it's an American company…

Do you think Chinese LLMs acquired training data legitimately? I think the whole situation is a bit funny, but I don't think the US "started it" to be fair.

> Do you think Chinese LLMs acquired training data legitimately?

I think they probably acquire it in accordance with Chinese law.

> but I don't think the US "started it" to be fair.

Who are you quoting with those marks? Started what? To be fair to whom?

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#144

Post by olejorgenb »

Tadpole9181 wrote: post #107
olejorgenb wrote: post #49

Earlier quoted context omitted.

The model is great. The UX is ~~horrible~~ annoying...

Don't get me started. For every half-decent choice, there's a multitude of insane choices. After all this time they still don't have side-by-side review. Equally as annoying, the break from VSCode is horrible . Having to use a separate registry, not having basic settings sync, the delay behind mainline VSCode updates. Then, it's just plain buggier than others. The agent terminal just doesn't work semi-regularly, it d…

Well, the UI as a whole is ok to me (except the parts which is way too volatile). I was talking about the UX of the autocomplete model. The model are very often spot on and fast, but it's impossible to properly configure it to be less in your face. Making it basically useless for day-to-day development.

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#145

Post by tempaccount420 »

lmc wrote: post #62
kbrkbr wrote: post #56

Earlier quoted context omitted.

Why not?

This 'Modified MIT' is not a license that has been through the OSI process: https://en.wikipedia.org/wiki/The_Open_Source_Definition#Com... You can't just add random terms to an existing license and use its name. "Modified MIT: Like MIT but pay us 50 million dollars." Perhaps CC-BY would've been more appropriate.

Ah yes, a document titled "*THE* Open Source Definition", describing *THEIR* definition of open source.

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#146

Post by charcircuit »

NitpickLawyer wrote: post #24
antirez wrote: post #20

Earlier quoted context omitted.

Basically this is true for most startups in the world BUT Cursor, so here you are kinda inverting the logic of the matter. Cursor is at a size that, if they wanted to use K2.5, they could clearly state that it was K2.5 or get a license to avoid saying it.

IF we assume that the modified MIT clause is enforceable. And if we assume Cursor Inc. is running the modification. It could very well be the case that Cursor Research LTD is doing the modifications and re-licensing it to Cursor Inc. That would make any clause in the modified MIT moot.

Now Cursor publicly claimed they didn't need to do anything since it was a partner provider that was serving the model and not them.

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#147

Post by dmix »

bearjaws wrote: post #81
dmix wrote: post #69

Earlier quoted context omitted.

Cursor’s integration is much deeper than just plugging an LLM into VSCode That said I have a feeling both VSCode and Claude code will catch up to their integration. But neither comes close yet (I say that as someone who mainly uses Claude Code).

As a command line junkie, what is the main thing Claude Code needs to catch up with cursor? I haven't dove into using a LLM in my editor, so I am less familiar with workflows there.

Mostly saying "include this line from x file and this block from y file" which keyboard shortcuts. Claude's VSCode plugin only does one selection. Claude Code requires explicitly telling it what to reference.

That plus Cursor's integration into VSCode feels very deep and part of the IDE, including how it indexes file efficiently and links to changed files, opens plans. Using Claude Code's VScode extension loads into a panel like a file which feels like a hack, not a dedicated sidebar. The output doesn't always properly link to files you can click on. Lots of small stuff like that which significantly improves the DX without swapping tabs or loading a terminal.

I also use Code from terminal sometimes but it feels very isolated unless you're vibecoding something new. I also tried others: Zed is only like 50% of the way there (or less). I also tried to use (Neo)Vim again and it's also nowhere close, probably 25% of the UX of Cursor even with experimental plugins/terminal setups.

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#148

Post by faangguyindia »

mohsen1 wrote: post #22

Cursor Composer 1 was Qwen and this is Kimi. IDE is based on VSCode. The entire company is build on packaging open source and reselling it. Ollama is also doing this. There is so much money to be made repackaging open source these days. So funny to see Twitter go wild saying "a 50 person team just beat Anthropic" blah blah.

It just means Kursor is sharing data with Chinese llm which enables them to improve their LLM by training on outputs and input of all data which cursor collects.

It's a two way street.

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#149

Post by nreece »

Partnership confirmed by Moonshot:

Cursor accesses Kimi-k2.5 via FireworksAI_HQ hosted RL and inference platform as part of an authorized commercial partnership.

https://x.com/Kimi_Moonshot/status/2035074972943831491

Cursor teams take:

Only ~1/4 of the compute spent on the final model came from the base, the rest is from our training. This is why evals are very different.

https://x.com/leerob/status/2035035355364081694

Re: Cursor Composer 2 is just Kimi K2.5 with RL

#150

Post by physicles »

bearjaws wrote: post #81
dmix wrote: post #69

Earlier quoted context omitted.

Cursor’s integration is much deeper than just plugging an LLM into VSCode That said I have a feeling both VSCode and Claude code will catch up to their integration. But neither comes close yet (I say that as someone who mainly uses Claude Code).

As a command line junkie, what is the main thing Claude Code needs to catch up with cursor? I haven't dove into using a LLM in my editor, so I am less familiar with workflows there.

You’re not missing much.

I used Cursor for the second half of last year. If you’re hand-editing code, its autocomplete is super nice, basically like reading your mind.

But it turns out the people who say we’re moving to a world where programming is automated are pretty much right.

I switched to Claude Code about three weeks ago and haven’t looked back. Being CLI-first is just so much more powerful than IDE-first, because tons of work that isn’t just coding happens there. I use the VSCode extension in maybe 10% of my sessions when I want targeted edits.

So having a good autocomplete story like Cursor is either not useful, or anti-useful because it keeps you from getting your hands off the code.

Post reply on HN