The anti-Cursor sentiment here is baffling to me given how useful it is to me. I use it interactively and actively review everything it produces. I like how I can plan a feature and refine the plan before instructing the agent to implement it. Last I checked, vscode had none of those features. Do (seemingly most) people prefer Codex because it gives a greater degree of autonomy to the agents?
> I like how I can plan a feature and refine the plan before instructing the agent to implement it
You can do that with claude code, github copilot (built into vs code) and codex, in any of their IDE versions, plugins for other ides (jetbrains, vscode, anything else you care to name) and also, of course, the CLI versions of all of them. They're also integrated into github, jira, and everything else.
Seriously, try other tools! if only to get a more balanced perspective.
This all being said, its been a long time since I last tried cursor... I'll give it a go.
I am personally not a fan of VS Code regardless, but I guess I don’t understand what it buys you over one code editor window and a Codex window both being open?
I have, right now, a tmux session with Codex on the bottom and Neovim on the top. It does what I was doing in Cursor just fine.
I am not really “anti Cursor”, I just genuinely am confused as to what it actually buys me over the setup I just described.
Here's why I use Cursor. My company pays for it, although I could switch to Claude Code or use Codex more since I also have ChatGPT enterprise account.
* Perhaps could be solved with the right terminal software, but I like the GUI for seeing my running agents and viewing all my conversations
* Works with multiple model providers in the same tool. I probably worry about cost optimization more than my employer would care for me to, but I frequently switch between openai/anthropic and switch between model sizes to use the tool that I think can get the job done for the least money. Another thing I like is having a long conversation with an expensive model, then I can switch to 5.4-nano to cheaply extract some little piece of information or summary from the conversation. Really this is big being able to switch model providers throughout the months without having to change my interface.
* Good support for the various ways of providing context. Rules, AGENTs.MD/CLAUDE.md files (if you want it to automatically read those), skills. Good hook support.
* I think the agent diff review experience is pretty good, but maybe it works similarly when you hook the cli agents into an editor, IDK.
* The default shell sandbox behavior is quite good. Every shell command runs in some sort of sandbox so that read only commands work without approval. The model asks for more permissions when it tries to do something that needs more permissions like network access or writing outside of the workspace directory. I know Claude code has a similar feature you can use.
* Good fork / revert conversation to checkpoints, with the option of reverting the code or just reverting the conversation.
* Feels decent that I am an API customer through Cursor. I don't hit Claude limits. Cursor doesn't have an incentive to limit reasoning or token usage, although they do have an opposite incentive.
* They are reasonably responsive to bugs and feature requests through their forum.
* Works well with a lot of repos / folders added to your workspace. I probably should organize all my stuff under a single directory, but alas I have like 8 different folders added to my workspace and it handles this well. Perhaps Claude --add-dir support works fine too.
DOWNSIDES:
* They are not quickly adding the best open source models to Cursor. Like Kimi 2.6 or whatever. Possibly not incentivized to given their Composer models.
* Don't love the subagent support. I can define custom subagents although it is not easy to get models to use mine instead of the builtin ones. The builtin ones do not allow me to control what model they run, so they will always run something like composer-2-fast, which is a fine model for all I know, but I would like to control it. Also, I would like if you could optionally make the subagent experience more first class. Like browse all the subagents and continue conversations with them or switch their model etc, although that is probably tricky / weird.
My employer pays for Cursor and Claude but not Codex. I often find Claude dumb (yes, even Opus), thus I'm using Cursor with GPT-5.4. If you have Codex, you don't miss anything.
NVIDIA has 42,000 employees. Even still, when their deal with Cursor comes to an end do we really expect them to stay loyal? And further, sign on with xAI?
When they could instead sign with the new hottest enterprise coding IDE (Claude, Codex, etc who are way more popular now). Maybe if it’s an acquihire, it’s the GTM/Sales that xAI is after?
They might. Elon will probably use his SpaceX/xAI spend (future SpaceX space datacenter dollars) as leverage. NVIDIA is so used to doing such deals by now, they’ll probably take it.
Space datacenters are such a dumb idea. We already have problems keeping spacecraft cool, and unless they send people up there with the datacenters who the heck is going to do maintenance? GPUs need upgrading and hard drives fail
> unless they send people up there with the datacenters who the heck is going to do maintenance
A fleet of special ships for orbital commute and future generation Optimus, no doubt.
OR - they follow the same approach as the current fleet of Starlink satellites. Probably have some redundancy baked in (eg 10 operational GPUs with 3 backup). They have rich data around hardware expectancy in space. Can’t hurt to have that for simulations.
at 42,000 employees and their own (infinite) compute on hand, there has to be at least one plucky junior internally who is suggesting using the open source equivalents, internal / open models and saving a big pile of money.
As someone put in the unfortunate position of building an AI IDE in three weeks, I assure you it's much more difficult than it seems. Sure, we were able to get something working with all the features you would expect, but the performance was awful. Claude Code, Cursor, and others do a lot of tweaking based on a lot of experience in order to make their systems give good results. There is more to getting good results than just using a good model.
