AUTO MODE IS ABSOLUTLY USELESS! But you know that already don’t ya!
@RoadkillRon could you create a bug report post with issues on Auto? I would like to pass the feedback to devs for improvements.
Output may vary depending on model but we want also to see what specifically is the issue so it can be addressed.
NO, It wont do any good! As previously stated, you already know all about auto and how “Well” it works. You already know its purpose! It’s all by design, what would a bug report do to change that?
@RoadkillRon really sorry to hear that. It’s not on purpose and there are parts of Auto that can be improved. Bug reports with Request IDs help with that and allow us to see what others experience in detail.
While general feedback is important bug reports are so much more potent in helping resolve issues. Even in the last days where there were complaints about bugs without specific details, those threads received more and more non specific similar replies by other users. At same time one user created a detailed report that allowed us to fix that bug as it was hard to reproduce without details.
Many are using Auto for simpler task like log checks, code analysis, looking for approaches,… and preserving Sonnet or similar models for complex tasks.
So I purchased an unlimited plan to check logs? I don’t remember that being part of the deal! I didn’t buy an unlimited plan to “Do simple tasks” I purchased an unlimited plan to do any task, as described in the deal we made originally. YOU CHANGED THAT DEAL WITHOUT MY CONSENT! Now you want me to help you fix the bugs LOL, I’m fine for now just educating the public with the truth for a while!
@RoadkillRon thank you for sharing your thoughts and feedback.
How are we supposed to use Claude Sonnet 4 for complex tasks when just a few prompts trigger a rate limit notice? Switching to Auto mode in the middle of a complex task is completely futile. On top of that, there are critical bugs in the latest update including one that deletes files . How is this even usable? I have stopped using it. Unfortunately, I still have about three weeks left in my billing cycle, but Cursor has become completely unusable.
1 Like
Let’s be honest please, most users in the forum already know that Auto is barely usable for anything beyond the most trivial tasks. It’s not just one person having a bad experience, this is a widespread sentiment echoed in thread after thread.
I’ve already stopped relying on it, but unfortunately, I still have a few weeks left on my billing cycle.
At this point, it really feels like the Cursor team is working extra hard to make sure everyone cancels their subscription. Between the broken features, pricing confusion, and lack of meaningful fixes, what’s the endgame here?
1 Like
@donxito You have engaged in similar discussions with team members in other threads already.
Criticism is totally fine and welcome. Spreading same discussions across the forum and insinuations are not ok.
If you would like me to look at an issue to improve Auto I am happy to do that with a Bug Report. Would also be glad to discuss how to do more with Auto in its own thread.
I hear you, and I’m not trying to cause trouble. But I also think it’s important to recognize that what’s happening in these threads isn’t just repetition, it’s momentum.
It’s not just me bringing this up repeatedly, it’s a growing number of users all experiencing the same issues with Auto and the pricing model. That’s not spreading, that’s the community reacting to a shared problem.
I totally understand the need for structure like bug reports, and I appreciate the offer to look into things. But let’s also be real, when key parts of the product are breaking, limiting, or regressing, people will speak up, and not just in one thread.
This isn’t about me vs. the team, it’s about longtime users feeling like they’re being priced out, slowed down, and ignored. Anyway I’m done
2 Likes
People should just request a refund and use a different service instead of forever complaining. Makes me wonder if they actually have work to do. There are lots of ai coding services out there. It’s clear Cursor has changed from what it was a year ago, but for people like me, it is still better than the competitors which seem to not be as interactive and contextual.
Has anyone found a worthy cursor competitor that has total ide integration, agents, fast tab completions, full project file awareness, and a restore checkpoint system? Most of the competition seems to be geared towards people vibe coding or needing entire projects setup via terminal. I need very specific problems dissected in my code with a lot of my guidance, and something that can spit out rudimentary code when necessary, this I think cursor does well at. I have tried having cursor’s “Auto” solve complex problems vaguely, and even if I choose sonnet manually it often still doesn’t just work, but it can get things close where I can work and understand it, identify problems, and then “Auto” is good enough to fix that. I think I would have to spend $100s a month to get an AI to entirely build the complex things I am building without my direct coding, which maybe I should do someday, but as you can see, that would be an entirely different price range and maybe a different service all together.
In short, Cursor seems best for developers who prefer or are required to be deeply aware of their code that are needing an assistant, not for developers who are looking to build out an entire product with some nudges and never code again. If you are developing something genuinely complex and unique (limited model training on), then AI is going to be very expensive to help you. But Cursor is a good assistant. You’ can’t beat this for $20/mo (or really $16). I have looked and haven’t found anything close, or I would switch.
People keep saying Claude Code is the solution. That is an entirely different product, and if that works for them, then it’s obvious why they are dissatisfied with Cursor. Simple google search shows people saying CC doesn’t have tab completions, but they don’t need them because they don’t even program anymore. Obviously what they are building with CC is not really that complex and there are probably 1000s of similar project on the internet that it trained on. So these developers are months away from being replaced if they are just “the prompt guy” or whatever they are building is going to be flooded with competition overnight making it pointless.
I think people are being a bit unreasonable and should simply switch services or be willing to pay a lot more money.
As long as unlimited Auto is not throttled and continues to give me Sonnet instead of gpt when the task demands it, I think this is a great product for developers who are still okay doing the programming.
4 Likes
