I’m a non-enterprise, solo developer using Cursor daily on a Django project.
My main issue isn’t prompt quality or AI behavior — it’s the growing mental overhead around model selection and usage limits. I used to be able to open Cursor and get work done. Now I spend time deciding:
what model to use for Plan mode,
what to switch to for implementation,
whether a task is a quick CSS change or deeper Django work that needs something else.
That decision-making didn’t used to be part of the workflow, and it adds stress. I feel like I need to understand AI models just to make Cursor usable throughout the month.
I like Cursor and want to keep using it. I just wish there were a simpler, more predictable default experience for non-enterprise developers who want to focus on building, not model management.
I do feel the same but unfortunately I think it’s just how it is. AI tech is evolving so rapidly, the models are changing and new ones are constantly being added. That’s completely out of Cursor’s hands and they’re just trying to keep up.
I do think there’s some peace of mind to be had in just sticking with old models that work well even though there are newer ones. I’ve been using GPT 4.1 heavily since it came out; it’s considered old and not that smart now, but for me it’s reliable and I’ve gotten very good at prompting it to get the results I want.
Recently (past month) I’ve mostly switched to either 5.1-codex-mini or 5.2-codex depending on the complexity of the task; again they both work well and codex mini helps save on costs. I haven’t tried any other models besides these three for many months now, because it’s just an endless rabbit hole where all the options change too fast for me.
Are you implying this amazing list of models is not clear for you ?
On a more serious note, I am feeling the same as you do. Having to reflect on the tool that we use more time than the project that we work on is not enjoyable.
I understand that you can pick whichever one you like, but right now it feels like I have to play a game of “which model does what I need but also doesn’t cost an arm and a leg.” Before, I could select auto and work for the entire month, focusing on my project… now that isn’t the case.
Yeah, this only became a headache once new AI models started dropping every week.
It’s like walking into a knife shop with a tight budget when all you need is something to cut cake. Suddenly you’re stressing about:
Will it actually work well?
Will it last more than a few uses?
Will the slices look clean or just mangled?
That’s AI right now. We want the perfect tool for every job, but unless you’ve got unlimited credits, that’s basically impossible. If money was no object, sure , grab the fanciest one, test everything, keep what feels best. But most of us have limits, so we gotta be smart about what we spend on.
My survival hack: pick 2–3 models, really get to know their quirks and strengths, then build little tricks / MCPs / prompts that make them way better for you. Choose your knives, then sharpen them until they feel like an extension of your hand.
One major reason confusion occurs is the **paradox of choice**. While having options is generally seen as freedom, too many choices can overwhelm the brain
Ideally DJ’s suggestion is absolute, choose a banger model and stick to it throughout. Eliminating the options will make your stress lesser. My approach recently is to use “Auto” mode everytime - it completely eliminates the thinking overhead of which model is good or bad. I’ve adviced against it but over a period of time i realised no model is going to stick around forever, so why bother model picking – let cursor do this for me.
As far as modes are concerned, i switch them according to usecase using “Shift + Tab” keys on the keyboard - trust me its better to build muscle memory for the key binding than even thinking of going and clicking and changing the mode.
Ask (by default) - I am constantly asking questions about my codebase, this is set as default. For small things, I don’t even ask the AI to write code, i just copy paste stuff from the Ask mode response. Think of this as READ mode.
Agent - THIS IS ALL YOU NEED. When cursor released, all they had is an agent and ask mode - the simplicity got a lot of traction for them. Think of this as WRITE mode.
The other modes are just for enthusiasts.
Ultimately, even if this is overwhelming and you want to use Cursor with NO STRESS at all - this is the formula:
如果不在意价格, Models | Cursor Docs 从定价里选择昂贵的模型,绝对不会错
而对于一般人的话,其实只要知道有gpt 和 claude 和 gemini这三个系列的模型,每个系列有大 小两种模型就够了,这样看的话其实就6个选项,每个都自己尝试一下不会很困难
而对于真正的玩家来说,新的AI模型像新的电脑硬件一样,每次新消息的发布都让人开心而不是信息过载
If price is not a concern, choosing the most expensive model from the pricing list will never be wrong. For the average person, it’s actually enough to know that there are three series of models: GPT, Claude, and Gemini, with each series having large and small variants. Looking at it this way, there are essentially six options, and trying each one yourself isn’t very difficult. For true enthusiasts, new AI models are like new computer hardware—every new release brings excitement rather than information overload.
