Cursor: When Pro-Developer Tools Chase the Magic App Creation Market

As a professional developer who's been using Cursor IDE since its early beta days, I've witnessed its transformation from an elegant tool built for developers to what feels increasingly like a product chasing the "anyone can code with AI" gold rush. This shift has left me, once one of their biggest advocates, searching for alternatives.

The Golden Days: Why Cursor Won Me Over

When I first discovered Cursor, it felt like someone had finally built the perfect AI-augmented IDE for professional developers. I immediately abandoned VS Code and never looked back. What made it so compelling?

Cursor respected my workflow. I decided what context the AI received, crafted precise prompts, and got back relevant code snippets that Cursor would automatically apply. This matched my mental model perfectly:

1. I selected the context (because I know my codebase best)
2. I formulated a clear, purposeful prompt
3. The AI responded with targeted changes
4. Cursor applied those changes seamlessly

While Cursor always had fancy code-search and auto-context features, I rarely used them. As most developers understand, code search and automatic context building are notoriously difficult problems. I preferred my manual approach because it gave me control.

The Pivot: From Developer Tool to "Magic App Creator"

Fast forward to today. Cursor has grown into a multi-billion-dollar startup with a very different focus. The company now appears to be targeting the much larger market of non-developers who want to create applications without writing code.

The signs of this shift are everywhere in the product:

Agent Mode Takes Over

Their new "agent mode" is now the default experience. Instead of responding directly to my queries, the AI now runs in autonomous loops, calling various tools and making decisions without my input. Often, it changes parts of my code I never asked it to touch.

I've watched in frustration as simple requests like "refactor this function to use async/await" turn into the agent diving deep into my codebase, rewriting unrelated functions, and suggesting architectural changes I never asked for.

Context Manipulation

Even more concerning, the "manual mode" I relied on has been compromised. Cursor now processes the context I select before sending it to the AI, apparently trying to optimize or reduce it through some opaque algorithm. The result? The AI no longer sees my files in full, leading to responses that miss critical context and suggest inappropriate changes.

Why This Matters: Tools Should Serve Their Users

I understand the business logic. The "create apps quickly with AI magic" market is larger than professional developer tools. Cursor has made a calculated decision to chase this opportunity.

But in doing so, they've killed the original product. The tool that once fit my workflow now feels like it's working against me, trying to be smarter than me rather than augmenting my capabilities.

A Broader Trend

This reflects a pattern in AI-powered development tools: starting with professional developers, then pivoting toward the more lucrative non-developer market. This creates problems for those of us who invested time adopting these tools, only to see them shift to serve a different audience.

What's Next?

I'm seeking alternatives that maintain what made early Cursor valuable: AI as an assistant rather than an agent, with the developer in control. I'm not against innovation, but the best developer tools respect their users' expertise.

If you're building AI tools for developers: we don't want magic. We want powerful, predictable tools that augment our capabilities while respecting our judgment.

What I'm building

Delegate tasks. Get software.

Give Vroni a GitHub issue, bug report, spec, or rough idea. It reads the repo, plans the change, writes code, runs checks, and works toward a review-ready pull request.

Take a look at vroni.com