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

Composer: Building a fast frontier model with RL

cursor.com

Post reply on HN
Print view View original

161–169 of 169 posts

Re: Composer: Building a fast frontier model with RL

#161

Post by DefineOutside »

pbowyer wrote: post #121
netcraft wrote: post #107

I love cursor, the tab completion and agent mode. But I really dislike vscode after using intellij for so many years. I really wish the underlying editor was better, or I could get cursor features in intellij instead. The editing of the files is mostly fine, but its everything else around it that a full IDE provides thats just so much better. Right now its intellij + claude code for me, and its fine, but I wish I cou…

Intellij's tab-complete is coming along; it's hit and miss if it will work but for similar edits I'm finding it picks up the pattern quickly and I can tab - tab - tab to make them happen. Still not up to Cursor standards though :)

I find Cursor's tab completion to be distracting enough with multi-line changes that I just disabled it, while I use IntelliJ's tab completion regularly.

Cursor's tab completion is better, but it doesn't seem to have a concept of not trying to tab complete. IntelliJ is correct half the time for completing the rest of the line and only suggests when it is somewhat confident in its answer.

Re: Composer: Building a fast frontier model with RL

#162

Post by pbowyer »

DefineOutside wrote: post #161
pbowyer wrote: post #121

Earlier quoted context omitted.

Intellij's tab-complete is coming along; it's hit and miss if it will work but for similar edits I'm finding it picks up the pattern quickly and I can tab - tab - tab to make them happen. Still not up to Cursor standards though :)

I find Cursor's tab completion to be distracting enough with multi-line changes that I just disabled it, while I use IntelliJ's tab completion regularly. Cursor's tab completion is better, but it doesn't seem to have a concept of not trying to tab complete. IntelliJ is correct half the time for completing the rest of the line and only suggests when it is somewhat confident in its answer.

I agree about the multi-line blocks Cursor proposes. Like it gets the first two lines right and then after that it's nonsense. I'd rather it stuck with a single line change at a time, and let me press enter before it predicts again.

Re: Composer: Building a fast frontier model with RL

#163

Post by vidarh »

mbrock wrote: post #152
vidarh wrote: post #124

Earlier quoted context omitted.

Some of these projects are at a "real business with a live site and users". Two of the current ones are. And it's not remotely "YOLO vibe coding". All the code gets reviewed, and tested thoroughly, and they are worked to specs, and gated by test suites. What I don't do is babysit the LLM until it's code passes both the test suite and automated review stages, because it's a waste of time. Others of these projects are…

To be clear, Claude probably also introduced the bugs? I tried to use an appropriate emoji to express the joking nature of this comment, but HN silently filtered it out, so pretend you see a grinning face.

No, Claude did not introduce the bugs. I caused the bugs, years ago, and didn't have time to pursue the project for a long time. Claude fixed them by being handed unfinished, broken code and a test suite and told to make the tests pass.

Re: Composer: Building a fast frontier model with RL

#164

Post by infecto »

diggan wrote: post #157
infecto wrote: post #155

Earlier quoted context omitted.

Disagree. The ultimate bar which is easily measurable, do users find value in it. Benchmarks are mostly meaningless especially in my opinion where cursor shines which is the tool chain. You can go try composer yourself today and see if it’s valuable to you.

Isn't that up to the reader/visitor/user to decide? As it stands right now, Cursor are publishing results they won't say how they got them, and compares them against aggregate scores we don't know the true results of, and you're saying "it doesn't matter, the tool is better anyways". Then why publish the obscured benchmarks in the first place then?

No I said I don’t believe any of the existing benchmarks do well when it comes to using a tool chain. They built a model specifically to be used with their tool chain calls, something that a lot of the models out there struggle with.

Re: Composer: Building a fast frontier model with RL

#165

Post by mbrock »

vidarh wrote: post #163
mbrock wrote: post #152

Earlier quoted context omitted.

To be clear, Claude probably also introduced the bugs? I tried to use an appropriate emoji to express the joking nature of this comment, but HN silently filtered it out, so pretend you see a grinning face.

No, Claude did not introduce the bugs. I caused the bugs, years ago, and didn't have time to pursue the project for a long time. Claude fixed them by being handed unfinished, broken code and a test suite and told to make the tests pass.

Ah, that's great. I've also found LLM agents extremely helpful for reviving old projects.

Re: Composer: Building a fast frontier model with RL

#166

Post by Sander_Marechal »

srush wrote: post #77
numbers wrote: post #53

Please keep the naming of your models sane, I'd like to know that composer 1 is the first model and composer 2 is second but composer 1o is not yet another 1 variant that's actually newer and better than 2, that's just dumb. Not that you're doing that, some other companies do that.

We will do our best. Luckily I don't think there are major telecom companies called Composer-2.

There is also a very polular package manager called Composer. Do companies not search for name collisions? Or do they squat on community projects on purpose?

Re: Composer: Building a fast frontier model with RL

#167

Post by W0WL0LXD »

chis wrote: post #72
srush wrote: post #64

Earlier quoted context omitted.

The blog talks about the training process. Specifically we trained with RL post-training on coding examples.

Makes sense, but what model was used for the base? Is it some open-source model, and you're not at liberty to disclose?

not a Cursor employee but still a researcher, it’s Zhipu/Z.ai GLM-4.6/4.5. there’s traces of Chinese in the reasoning output + its the only model that would make sense to do this with RL, and is a model that already delivers near SOTA performance + is open-source/open-weight.

Cursor Composer and Windsurf SWE 1.5 are both finetuned versions of GLM.

Re: Composer: Building a fast frontier model with RL

#168

Post by jasonjmcghee »

JanSt wrote: post #156
jasonjmcghee wrote: post #43

Earlier quoted context omitted.

Yup - just like sibling comment said - my "low bar" is going to be whatever the best model is that isn't unreasonably costly/expensive. Speed of model just isn't the bottleneck for me. Before it I used Opus 4.1, and before that Opus 4.0 and before that Sonnet 4.0 - which each have been getting slightly better. It's not like Sonnet 4.5 is some crazy step function improvement (but the speed over Opus is definitely nice…

I think Opus 4.1 is still much better than Sonnet 4.5

If cost is not considered- absolutely. That being said sonnet 4.5 and using thinking where it makes sense feels like way more bang for your buck and usually good enough. I really don't use opus anymore

Re: Composer: Building a fast frontier model with RL

#169

Post by chaidhat »

W0WL0LXD wrote: post #167
chis wrote: post #72

Earlier quoted context omitted.

Makes sense, but what model was used for the base? Is it some open-source model, and you're not at liberty to disclose?

not a Cursor employee but still a researcher, it’s Zhipu/Z.ai GLM-4.6/4.5. there’s traces of Chinese in the reasoning output + its the only model that would make sense to do this with RL, and is a model that already delivers near SOTA performance + is open-source/open-weight. Cursor Composer and Windsurf SWE 1.5 are both finetuned versions of GLM.

interesting, thank you
Post reply on HN