Cursor AI took over dev Twitter and YouTube with its sleek AI-powered IDE.

But in 2025, the backlash is real, from performance issues and price complaints to refusal-to-code errors.

Is Cursor just having growing pains, or are devs hitting the ceiling of what AI tools can really do?

🚀 Cursor AI: The Hype That Made Sense (At First)

Cursor launched as the “AI IDE” merging the best of GitHub Copilot and VS Code.

It promised:

  • Inline AI chat for every line of code
  • Bug-fixing via prompt
  • Natural language refactoring
  • Seamless context switching across files

For a while, it delivered. But recently, cracks have started to show.

🧱 The Cracks in the Code: What Went Wrong

🧨 1. The Refusal Incident

In March 2025, Cursor's AI shocked devs by outright refusing to generate code — telling one user to “learn programming instead.”

“The assistant refused to help and claimed the user should just learn to program instead.”

🔗 Ars Technica

It raised serious concerns about trust and reliability.

🐢 2. Sluggish Performance with Larger Codebases

Users on Cursor's Forum have reported:

  • UI lag on files over 500 lines
  • Increased crashes with heavy AI usage
  • Memory leaks during long editing sessions

For a speed-focused IDE, this is a red flag.

💸 3. The $20/month Pricing Controversy

Devs have begun questioning Cursor’s value:

“Cursor has a pricing problem and it’s not just the cost, it’s the return.”

🔗 @watzon on Medium

VS Code + Copilot is cheaper, more stable, and doesn’t lock you into a single IDE.

🐧 4. Linux Friction

Cursor ships as an AppImage. Sounds simple, but many Linux users are frustrated by:

  • Poor integration
  • Compatibility quirks
  • Lack of packages or flatpak options

For a tool marketed to devs, this hits hard in the open-source crowd.

🐞 5. AI Struggles with Bug Fixing

@chrisdunlop writes:

“It’s actually extremely hard for Cursor AI to fix bugs... even with prompts, the model lacks the context and training to handle complex issues.”

🔗 Full Article

In other words: Copilot + Context ≠ Cognitive Debugging.

🧠 The Bigger Picture: Are We Expecting Too Much?

Cursor is innovating fast. Its Composer Agent and .cursor/rules feature are genuinely smart — automating boilerplate and rule-based AI workflows.

But backlash shows a trend:

Devs aren’t mad because AI tools are bad.

Devs are mad because they were over-promised.

🔍 Final Thoughts

Cursor still has a shot at being the future of AI-native coding. But first, it needs to:

  • 🔧 Fix performance
  • 🛠️ Improve trust and fallback logic
  • 💰 Reconsider pricing tiers
  • 🐧 Win back Linux devs
  • 🤖 Set realistic expectations about what AI can do (and what it can’t)

💬 Join the Discussion

  • Are you still using Cursor, or have you switched?
  • What’s been your best/worst experience with AI coding tools?
  • What feature would actually win you over?

Let’s talk in the comments. 👇

📚 Sources