Quick reply while I have a sec! Assistants usually lock you into a fixed $20 sub regardless of usage. That matters cuz you end up overpaying if youre not coding 24/7. For your situation, I suggest using OpenAI API keys with the Continue extension for VS Code. Its way more precise and you only pay for tokens used for React tests. Best budget setup tbh. gl!

hey! I would suggest comparing GitHub Copilot vs Cursor Code Editor. cursor is AMAZING for React!! Copilot is okay for snippets, but Cursor Code Editor basically understands your whole folder structure... makes unit tests sooo much easier. i think ChatGPT Plus is still pretty worth it for Custom GPTs too? maybe? but Cursor is highkey the best choice for a smooth workflow tho!! gl

To add to the point above: I really think we need to be more careful with how much autonomy we give these tools. Stumbled upon this discussion and felt like I had to chime in because I've been burned before. A few months back, I let an automated tool refactor a complex React component tree, and it felt like magic... until I realized it had completely broken the memoization logic. It introduced a subtle memory leak that took me days to track down. My current setup is much more conservative now. I mostly use a simple chat interface and manually copy bits over because I need to vet every single line. It might feel slower than the one-click solutions people are talking about here, but for me, reliability is everything. You definitely want to make sure you're not sacrificing long-term stability for short-term speed. I treat the AI like a rubber duck that occasionally talks back—useful for ideas, but I dont let it touch my file system directly anymore. It's just not worth the risk when you're dealing with complex logic.

Helpful thread 👍

I get the hype around those automated editors, but ngl, I’m always a bit wary about letting an AI have full reign over a whole project structure. While they're fast, I’ve seen them introduce some really weird, subtle bugs in complex React logic that are a nightmare to debug. For me, reliability is everything, especially when you're dealing with state management or lifecycle stuff. Instead of a full editor replacement, I’ve had a lot of success using the Aider command-line tool. It’s a bit more technical, but it feels way safer because it shows you exactly what it's changing in a git-like format before anything is finalized. Also, for unit tests, I usually stick to a custom GPT I made specifically for React Testing Library best practices. I basically treat it like a pair programmer that I have to double-check constantly. Iirc, there's also Sourcery which is decent for refactoring without the hallucination risk you sometimes get with the bigger LLMs. Just my two cents, but staying in control of the changes is huge for keeping things stable (at least that's what worked for me).

Works great for me