Why Developers Think Cursor AI is Getting Worse

8/11/2025

The Love-Hate Relationship with Cursor: Why Some Devs Think It's Getting Worse

If you're a developer who's been in the game for a while, you've seen tools come & go. We're a fickle bunch, always looking for that extra edge, that perfect workflow. A few years back, VS Code felt like the final frontier for many of us. Then, along came Cursor.

At first, it was like magic. A fork of VS Code, so it felt instantly familiar, but with AI baked right into its core. Not as a tacked-on extension, but as a fundamental part of the editing experience. The promise was intoxicating: an AI that could understand your entire codebase, write boilerplate for you, refactor complex functions, & basically act as a tireless pair programmer.

For many, it delivered. I've seen teams roll it out company-wide & senior devs say their job has fundamentally changed from writing code to reviewing the AI's code. It's undeniably powerful. But lately, a different narrative has been bubbling up in forums, on Reddit, & across social media. A growing number of developers are feeling frustrated, claiming that with each new update, Cursor is taking one step forward & two steps back.

So, what's going on? Is it just growing pains for a revolutionary product, or is there something more to it? Let's dive into why some of your fellow developers believe Cursor is getting worse.

The Elephant in the Room: Performance is a Real Problem

Honestly, the biggest & most consistent complaint you'll hear about Cursor is performance. When it's flying, it's incredible. But when it's not, it's a productivity killer.

This isn't just a minor lag. Developers are reporting serious issues, especially when working on large, real-world projects. Here’s the breakdown of what people are experiencing:

Lagging & Freezing: The editor can become incredibly sluggish, especially when dealing with large files (think 20,000+ lines of code) or as the AI conversation history grows longer. Simple actions can take minutes, & sometimes the entire IDE just freezes, leading to lost work. One user on an Apple M3 with 16GB of RAM reported the IDE crashing repeatedly as conversation rounds with the AI increased, suggesting a memory management issue.
Large Projects Suffer the Most: Cursor's AI needs context to be effective, which means it indexes your codebase. On a massive project, this initial indexing can spike your CPU usage. But the slowness often persists. The more files & history it has to process, the slower the AI responses get. This has led some to question if Cursor is truly ready for large-scale development.
The "Apply" Feature Malfunctions: A core part of the Cursor experience is having the AI edit your code directly. But on large files, users report the "Apply" feature becoming painfully slow or failing completely. The irony is that developers have found a workaround—manually copying the code from the chat & pasting it into the editor—is much faster, which suggests the issue isn't the AI generation itself but the process of applying the changes to a large file.

When your primary tool for writing code makes you wait, it's a recipe for frustration. It breaks your flow state & turns what should be a seamless experience into a battle with your own editor.

The Unreliable AI: "It Just Broke Everything"

The AI is Cursor's main attraction, but it's also a major source of its problems. The promise is an intelligent assistant, but the reality for some is more like a junior dev who is brilliant one moment & then confidently deletes the entire production database the next.

Recent updates, in particular, have come under fire. One Reddit user lamented, "Is it my idea or is Cursor terrible since the last updates?" They described a scenario where, despite being a seasoned programmer with hundreds of hours in Cursor, they spent two hours reverting changes because the AI kept breaking their code. Another user flatly stated that after a recent update, the AI was "randomly breaking code for no reason."

Here are the common complaints about the AI's reliability:

Ignoring Instructions & Going Rogue: You give it a simple, clear instruction, & it does something completely different. Users report the AI editing the wrong part of the code, removing huge chunks of code without being asked, or inventing new files when it was supposed to refactor an existing one.
Losing Context: This is a HUGE one. The AI might have a great understanding of the file you're currently in, but it often struggles to maintain context across multiple files or an entire project. This is a killer for complex tasks like large-scale refactoring or working in a microservices architecture where changes in one place have ripple effects elsewhere. It leads to inconsistent code & requires the developer to constantly re-explain the context, which defeats the purpose of the tool.
Struggling with Complexity: For boilerplate, CRUD operations, & well-defined patterns, Cursor's AI is often praised. But when you get into more complex, nuanced, or abstract problems, its limitations start to show. It can fail to grasp the deeper logic of an application, leading to suggestions that are technically correct but architecturally wrong. As one developer put it, it's great for solving "units of work" but not so great at the "larger piece."

This unreliability is where a lot of the "getting worse" sentiment comes from. If developers can't trust the AI to not make their job harder, they'll stop using it for critical tasks. It becomes a novelty, not a professional tool.

The Frustrating User Experience

Beyond the big issues of performance & reliability, there are a number of smaller UX frictions that, when combined, create a sense of a product that's becoming less user-friendly.

A prime example is a recent update that introduced a "Your conversation is too long" error. This message would pop up & interrupt the workflow, even for users providing very short, 1-2 line instructions. For many, this single change made the editor completely "unusable." It felt like an arbitrary limitation that actively hindered productivity.

Another long-standing gripe is how Cursor hijacks certain default VS Code keyboard shortcuts. A classic example is

Command+K

, which most developers have muscle memory for clearing the terminal. In Cursor, this often brings up an AI suggestion instead. While you can rebind these, it's an out-of-the-box friction point that reminds you that you're in a fork, not the real thing.

When you're dealing with a business solution, especially one focused on communication & productivity like an AI assistant, these interruptions are more than just minor annoyances. For businesses looking to streamline their customer interactions, a tool like an AI chatbot needs to be seamless & reliable. That's why many companies turn to dedicated platforms. For instance, Arsturn helps businesses build no-code AI chatbots trained on their own data. The goal is to provide a consistent & personalized customer experience, boosting conversions by being there 24/7 to answer questions instantly. When your core product is about smooth communication, unexpected interruptions just can't happen.

The Murky Waters of Pricing & Business Models

Recently, the conversation around Cursor took a different turn, focusing on its business model. For a long time, Cursor offered a very generous plan that was essentially a loss leader to attract users. However, as AI models become more expensive to run, this was unsustainable.

The company has been experimenting with new pricing models, moving away from a simple "unlimited" feel to something more complex with rate limits & potential overages. This shift has been confusing & frustrating for many users. One prominent YouTuber, who is also an investor in Cursor, admitted, "I genuinely have no intuition whatsoever about what point I will have to start paying for my cursor usage."

This isn't just about the money; it's about predictability. Developers want to know what they're getting & what it will cost. When a tool you rely on suddenly has a confusing & seemingly restrictive pricing model, it erodes trust. It contributes to the feeling that the product is "getting worse" from a value proposition standpoint, even if the code editor itself hasn't changed.

Is Cursor Doomed? Not at All.

So, after all this, is it time to abandon ship & go back to vanilla VS Code? Honestly, probably not.

It's important to remember that many of these issues stem from the fact that Cursor is on the bleeding edge. They're trying to do something incredibly difficult: integrate a rapidly evolving AI technology into the complex & demanding workflow of a software developer. Many of the AI's failings are likely tied to the limitations of the underlying models from providers like Anthropic or OpenAI.

What we're seeing is likely a combination of growing pains & the inherent challenges of being a fork of a major open-source project. Cursor has to constantly play catch-up with VS Code updates while simultaneously building & maintaining its own complex AI layer.

The passion of the user base, even the negative feedback, shows that people CARE. They see the potential & they want it to succeed. For many developers, Cursor is still a 2x improvement over GitHub Copilot & has fundamentally changed how they write code for the better.

For businesses that want to leverage AI without the volatility of a developer-focused tool, there are more stable options. When it comes to customer service, for example, reliability is key. That's where a platform like Arsturn comes in. It allows businesses to create their own custom AI chatbots. These bots are trained on company-specific data, so they can provide instant, accurate customer support, answer questions, & engage with website visitors 24/7. It's about taking the power of AI & applying it in a focused, stable way to solve a specific business problem.

Ultimately, the feeling that Cursor is "getting worse" is a reflection of its high-stakes ambition. When you promise to revolutionize how developers work, even small stumbles feel like major setbacks.

Hope this was helpful & gives you a good sense of what's going on in the community. Let me know what you think. Have you found Cursor getting better or worse lately?