I'm in the same boat, about to get stuck in to Cursor. But I'm wondering if it might be better to go with a VS Code extension like Continue.

With Continue you can use your own local copy of an LLM, and also query LLM API endpoints, so therefore better for a pay as you go solution and cheaper if you are a heavy user. Also you can carry on using VS Code as your editor of course.

This is my understanding of the trade offs. I assume that Cursor has better functionality and is better at coding for it to be a serious option compared with the VSCode extensions. I'm going to try out both and see what I like best.

My experience, and also what I’ve seen from a few YT videos as well, is that the local LLMs just aren’t as good as Claude and chatGPT. They don’t seem to understand or follow instructions as well as C and C.

Yes, good point. The latest LLMs from Anthropic, Open AI etc will always be more powerful than your local LLM. I reckon that the cutting edge cloud-based models will (continue to) provide enormous value for a range of complex actions, but that we will also derive great value from our own personal and organizational local models running on our devices and LANs.

I'm assuming that small specialized LLMs for coding will become 'good enough' for a significant proportion of tasks. For example, when Codestral came out earlier this year it supposedly beat Meta’s CodeLlama 70B model in all of the top coding evaluation benchmarks, even though it's 3 x smaller.

It looks like it takes three different models chat/ranking/ etc. What is the total cost for three API or VRAM requirements for reasonable local open source models to run Continue?

Just set it up 2 days ago. I fully agree with his take. Cursor does a lot to reduce the headache of copy pasting results from ChatGPT/Claude.

Makes coding with AI much faster and less of a chore. You still need to know what you're doing as the models are fantastic at boilerplate but often get things wrong. But it's significantly sped up iteration for me, and hence translates to more fun.

What specific features make Cursor better than GitHub Copilot?

Did anyone have similar experience with Cody comparing cody with cursor? Most of the metrics does not correlate much with everyday programming tasks.

Excluding sourcegrah employees who usually comment on these comparisons comments.

Wondering the same, I've been using Cody and after looking at these Cursor videos am not really seeing what's better about it (not affiliated to Sourcegraph). Just looks very similar but you have to use their VSCode fork. Being able to use a local model could become an advantage but not relevant when talking about using it with Sonnet 3.5.

I'm not using Cursor, but I'm still using Sonnet 3.5, although with https://aider.chat/ which is basically a TUI program to make the LLM a "pair programmer" - it can replace chunks of code, operate over multiple files, there's a repository map for it to check (for symbols), etc.

I haven't tried Cursor, but I've been messing around with Aider (aider.chat) with interesting results. With it, you add files to its context and then describe the edits you want. It then writes a commit with (hopefully) those changes. You can use it with pretty much any model, though only close-to-SOTA models work well. I've had pretty good results with Deepseek Coder, which is tens of times cheaper than Claude.

I'm using Zed alongside its new LLM integration with Claude Sonnet 3.5 and I'm really really enjoying it. The workflow and the way you control what goes into the context feels very good and efficient and mimics the way I've been using LLMs before, just MUCH faster and convenient.

Likewise, I have been doing the same.

I have an idea for a small side-project and I have been trying this workflow for styling via tailwind, adding some interactivity, etc. It can get surprisingly good, especially by easily sharing terminal output, access to files and folders and other convenient things.

Definitely worth checking out!

I just started using it a few days ago after hearing about it here.

So far, it feels transformational for my productivity. I've been using the ChatGPT and Claude webapps a few times a week for specific problems for a year, but this level of integration makes a big difference. Particularly for parts of the stack I'm not super experienced with -- I'm a backend/ML developer, but I was putting together a web UI with Tailwind for a personal project yesterday. I highlighted the ~20 lines of HTML for a menu bar, pressed Cmd+K and typed "can you make this take up a little less space on mobile?" and it adjusted the breakpoints and sizes of multiple elements perfectly. Pretty cool stuff.

Why was this nuked from the front page? I was really looking forward to this discussion.

I didn’t flag, but it reads a bit like promotion framed as a question to me.

If users flag a story on hn, some kind of ranking penalty is applied.

Few months in with Cursor as my daily driver and I completely agree. I cannot unsee my productivity gains.

JetBrains has a pretty good integration with CoPilot, am wondering how cursor compares with it.

It's been pretty good so far, settled as my main code editor. Prior to cursor I had to constantly provide context to GPT/Claude, now that's not necessary with cursor.

It's nice. Look into the composer. I like it much better than inline completions.