Adding my two cents, I had issues with standard plugins lately since they arent as good as expected for complex refactors. unfortunately, GitHub Copilot Individual highkey let me down, so I recommend Sourcegraph Cody because it actually indexes your codebase via embeddings. it wont break your logic like the others. gl!

> I've been using GitHub Copilot, but I'm curious if newer tools like Cursor... are better now. For your situation, I gotta say Cursor AI Code Editor is the real deal. Most tools are just plugins, but deep integration matters cuz it actually understands context. I moved from Copilot and honestly, its sooo much better because it indexes ur whole codebase. Its reallyyy worth the $20/month Pro plan, totally pays for itself in saved time!! basically a game changer, right?

Man I wish I found this thread sooner. Would have saved me so much hassle.

Can vouch for this

Before I give advice, can you clarify what ur actual monthly budget is? Ngl, costs add up SO fast. Just a few warnings first: - privacy is a HUGE issue with some newer startups
- some tools literally just hallucinate logic during refactors
- subscriptions bleed you dry if ur not careful Anyway, what stack are you usually working in??

Been using this for years, no complaints

Interested in this too

Ok so, I'm still kinda new to this whole AI dev world, but honestly I've been super cautious about what I let touch my codebase. Basically, you gotta remember that these LLMs are just guessing logic, so why it matters is that they can hallucinate security holes without you even noticing!! It's lowkey scary if you're not careful. Here's what I recommend if you want to be safe: * Tabnine Pro: I like this because it offers local deployments. It basically keeps ur code private on ur own machine, which is a HUGE plus for security. The downside is it might not be quite as 'magical' as the cloud ones, but it's reliable.
* Amazon CodeWhisperer Individual: This is great because it actually has built-in security scanning. It's really good at suggesting code that follows best practices, tho it's definitely most powerful if ur already in the AWS ecosystem. I mean, just make sure to double-check everything... dont just blindly trust the AI lol. Good luck with the workflow!! peace

Ok so, I've been obsessed with looking at actual performance benchmarks lately because honestly, if the latency is high, it just kills my flow!!! Tbh, you should really focus on the tools that prioritize raw speed and efficient context window management above everything else right now. Here is the direction I’d go for real-world performance:
* Just get the AI assistant stuff from JetBrains, it's basically the most stable since it’s baked into the IDE.
* Look into any of the newer 'ultra-low latency' autocomplete brands if you need that instant response time.
* Try out the open-source agentic platforms if you want to run your own local tests and keep everything snappy. I'm not 100% sure which one will win the long-term benchmark wars, but going with a performance-first provider is the only way to scale without getting frustrated by lag? If it takes more than a second to think, it's basically breaking your focus anyway.

Same boat, watching this

This is exactly what I needed to hear. Youre a lifesaver honestly.