As the other commenter stated, I don't use CLIs for development. I use VSCode.

I'm very pro IDE. I've built up an entire collection of VSCode extensions and workflows for programming, building, customizing build & debugging embedded systems within VSCode. But I still prefer CLI based AI (when talking about an agent to the IDE version).

> Composer 1

My bet is their model doesn't realistically compare to any of the frontier models. And even if it did, it would become outdated very quickly.

It seems somewhat clear (at least to me) that economics of scale heavily favor AI model development. Spend billions making massive models that are unusable due to cost and speed and distill their knowledge + fine tune them for stuff like tools. Generalists are better than specialists. You make one big model and produce 5 models that are SOTA in 5 different domains. Cursor can't do that realistically.

▲

santoriv 3 days ago | parent | next [-]

> My bet is their model doesn't realistically compare to any of the frontier models.

I've been using composer-1 in Cursor for a few weeks and also switching back and forth between it, Gemini Flash 3, Claude Opus 4.5, Claude Sonnet 4.5 and GPT 5.2.

And you're right it's not comparable. It's about the same quality of code output of the aforementioned models but about 4x as fast. Which enables a qualitatively different workflow for me where instead of me spending a bunch of time waiting on the model, the model is waiting on me to catch up with its outputs. After using composer-1, it feels painful to switch back to other models.

I work in a larg(ish) enterprise codebase. I spend a lot of time asking it questions about the codebase and then making small incremental changes. So it works very well for my particular workflow.

Other people use CLI and remote agents and that sort of thing and that's not really my workflow so other models might work better for other people.

▲

neutronicus 3 days ago | parent [-]

Does it have some huge context window? Or is it really good at grep?

The Copilot version of this is just fucking terrible at suggesting anything remotely useful about our codebase.

I've had reasonable success just sticking single giant functions into context and asking Sonnet 4.5 targeted questions (is anything in this function modifying X, does this function appear to be doing Y) as a shortcut for reading through the whole thing or scattershot text search.

When I try to give it a whole file I actually hit single-query token limits.

But that's very "opt-in" on my part, and different from how I understand Cursor to work.

▲

santoriv 3 days ago | parent [-]

It is really good at grep and will make multiple grep calls in parallel.

And when I open it in the parent directory of a bunch of repos in our codebase, it can very quickly trace data flow through a bunch of different services. It will tell me all the files the data goes through.

It's context window is "only" 200k tokens. When it gets near 200k, it compresses the conversation and starts a new conversation..... which mostly works but sometimes it has a bit of amnesia if you have a really long running conversation on something.

▲

Yoric 2 days ago | parent [-]

> It is really good at grep and will make multiple grep calls in parallel.

How does that work? Multiple agents grepping simultaneously?

▲

dimava 2 days ago | parent | next [-]

When other models would grep, then read results, then use search, then read results, then read 100 lines from a file, then read results, Composer 1 is trained to grep AND search AND read in one round trip It may read 15 files, and then make small edits in all 15 files at once

▲

simondotau 2 days ago | parent | prev | next [-]

Presumably if it knows it needs to perform multiple searches in order to gather information (e.g. searching for redundant implementations of an algorithm, plus calls to the codebase's canonical implementation) it should be able to run those searches in parallel grep calls.

	▲	Yoric 2 days ago \| parent [-]
		I'm trying to figure that one out. LLMs are inherently single-threaded in how they ingest and produce info. So, as far as I can gather from the description, either it spawns sub-agents, or it has a tool dedicated for the job.

▲

void-star 2 days ago | parent | prev [-]

Probably something closer to ripgrep, if not actually ripgrep.

▲

spruce_tips 3 days ago | parent | prev [-]

composer 1 has been my most used model the past few months. but i only use it to execute plans that i write with the help of larger, more intelligent models like opus 4.5. composer 1 is great at following plan instructions so after some careful time providing the right context and building a plan, it basically never messes up the implementation. sometimes requires a few small tweaks around the edges but overall a fantastic workflow that's so delightfully fast