Which AI tools are ...
 
Notifications
Clear all

Which AI tools are best for software developers today?

10 Posts
11 Users
0 Reactions
299 Views
0
Topic starter

Hey everyone! I’ve been seeing so many new AI tools popping up lately that it’s getting hard to keep track of what’s actually useful for a daily coding workflow. I’m currently using GitHub Copilot, but I’ve been hearing a lot of buzz about Cursor and Claude 3.5 Sonnet for better refactoring and logic handling. I’m really looking for tools that can help with complex debugging or generating unit tests without constant hallucinations. I want to streamline my process, but I’m worried about picking tools that add more overhead than they save. Are there any specific IDE extensions or AI assistants you guys swear by for full-stack development right now?


10 Answers
11

Ok so I went through this last year when my subscription costs were spiraling out of control! I've been coding for a decade and honestly, the tool fatigue is REAL. I basically switched to a *pay-as-you-go* workflow to save cash.

1. I started using the Anthropic Claude 3.5 Sonnet API through a lightweight wrapper. It literally cost me like $5 last month instead of $20.
2. For unit tests, I've had zero hallucinations using Cody by Sourcegraph Free Tier which is actually fantastic for context-aware code.
3. I also tried the Tabnine Basic Plan because its free and runs locally, so no privacy concerns!

It definitely took a minute to setup, but saving $200+ a year is amazing! gl!


10

So basically the consensus is that Cursor AI Code Editor paired with Claude 3.5 Sonnet via the Anthropic Claude API is the current king for logic and refactoring. Honestly, I was a die-hard GitHub Copilot fan for a year, but after trying the other setups mentioned here, I realized how much market competition has heated up!

Looking at the market right now, here is why everyone is switching:
- **Context Awareness**: Cursor AI Code Editor actually indexes your whole folder, which helps with those complex debugging sessions you mentioned.
- **Model Flexibility**: Unlike Copilot, these newer tools let you swap between models like GPT-4o or Claude depending on the task.
- **Cost Efficiency**: Using the Anthropic Claude API with a pay-as-you-go model is highkey smarter if you aren't coding 24/7.

TL;DR: The thread definitely leans towards Cursor. It's SO much better at generating unit tests without the usual hallucinations. I'd say give it a shot for a week and see if it sticks! peace


3

Good to know!


2

Did this last week, worked perfectly


2

Seconded!


1

Honestly, i feel u on the tool fatigue. I was sticking with the basics for a while, but lately I've just been using Cursor and it's been a total game changer for my workflow. It handles the logic stuff sooo much better than my old setup. I basically just let it handle the unit tests now and it's actually reliable. Seriously, just go with any of the newer AI-native IDEs and ur gonna be happy.


1

Seconding the recommendation above! Honestly, Cursor AI Code Editor is such a game changer, especially since it lets you toggle between models. But if youre worried about budget like I am, I gotta mention Claude 3.5 Sonnet via the Anthropic API. Instead of a flat $20/month sub, you can just pay for what you actually use through their console, which is AMAZING for saving a few bucks when youre not coding 24/7.

I also recently tried the Codeium Individual Plan cuz it has a free tier that is actually legit for full-stack stuff! It helps with that logic handling you mentioned without the crazy overhead. Ngl, I am still kinda a beginner with all the complex debugging stuff, but using the API credits feels way more cost-effective than subscribing to every single tool out there. Definitely check out the pay-as-you-go options if you wanna keep costs low!! hope that helps haha


1

Following


1

@Reply #5 - good point! Simplicity is usually the way to go to avoid tool burnout. Over the years I've tried many setups, and honestly, the performance of Supermaven Pro has blown me away lately. Its context window is massive, which is great for large repos, but the real seller is the zero latency. It doesn't lag like some of the heavier plugins that try to do too much at once. If you're cautious about privacy and want to avoid that overhead feel, I've been using the Continue.dev Extension for VS Code quite a bit. It is open-source and basically lets you pipe in whatever model you trust most for debugging. In my experience, using it with Mistral Large 2 provides much more grounded logic for unit tests compared to the more creative models. It takes a tiny bit of configuration, but it helps streamline the process since you arent constantly fighting the AI to stay on track. Speed and reliability are everything when you're deep in the logic... just gotta find what fits your specific rhythm.


1

Ok adding this to my list of things to try. Thanks for the tip!


Share: