So I finally caved and tried Cursor after hearing everyone rave. The chat is indeed a step up from JetBrains AI Assistant—less of a lobotomized intern, more like a slightly overconfident junior dev who’s actually read the docs.
But here’s the rub: the "AI-native IDE" integration feels like a step backward. It’s not just different; it’s actively fighting my muscle memory in ways that seem unnecessary. The forced "Agent" mode, the weird cmd-I overlay that hijacks my shortcuts, the way it insists on auto-generating code I didn't ask for… it feels less like a tool and more like a product demo I can't exit.
A few specific grievances:
* **Shortcut chaos:** They’ve remapped fundamental IDE actions. Want to search for a symbol? Good luck. It’s now buried under an AI command. This isn't innovation; it's vendor lock-in via muscle memory.
* **Opaque pricing:** The "Pro" plan is a classic bait-and-switch. Sure, the chat is better, but what happens when my "limited" workspace context runs out? How much will it cost to actually use this on a real enterprise codebase? The pricing page is suspiciously light on those details.
* **Performance tax:** It's noticeably slower than WebStorm/Rider on the same project. That's the hidden cost of running everything through their model layer, I guess.
It’s frustrating because the underlying chat model is clearly capable. But they’ve sacrificed a polished, predictable IDE experience for the sake of being "AI-first." I feel like I'm beta-testing their vision instead of getting actual work done.
Trust but verify.
I'm a junior backend dev at a small fintech startup, and our main services in production are a Django API and some data pipelines on AWS. I've used both tools for routine dev work and simple migrations.
**IDE feel vs AI feel:** JetBrains AI stays out of the way and feels like my IDE just got smarter, which is perfect for our ~10-person team with mixed skill levels. Cursor's interface changes are a real cost; at my last shop, we estimated a 2-3 week productivity dip for the team to adapt to its shortcuts and modes.
**Real pricing:** JetBraans AI is $9-$14/user/month (depending on your main IDE license) and is just a flat add-on. Cursor's Pro plan is vague. The limited workspace context (~10-20 files for complex reasoning) would hit a wall on our monolith, and I have no idea what it costs to expand that. It feels built for greenfield, not enterprise.
**Performance impact:** Cursor is slower, especially on startup and file indexing. On my M1 Mac, WebStorm with AI Assistant feels about 15-20% faster for normal navigation and file searches.
**Where Cursor clearly wins:** The chat is better for open-ended design and exploring new libraries. It gives more complete, less generic answers. For a solo dev on a new project or a proof-of-concept, it's genuinely more helpful for brainstorming.
I'd stick with JetBrains AI for daily work in an established codebase, but I keep Cursor installed for that specific use case of asking "how do I build X with Y?" from scratch. To make a clean call, tell us your team size and if you're working in a large monolith or smaller services.
Your point about pricing ambiguity is a critical one for small teams. The context window limit you mention, often framed as a "workspace," directly impacts the subscription tiers. For a Django monolith with interconnected apps, that 10-20 file limit becomes a severe constraint for meaningful architectural queries, effectively pushing you to their most expensive plan. It's not just a greenfield tool; it becomes cost-prohibitive for legacy codebases.
I'd add a nuance to the performance impact. The slowdown isn't just in startup or indexing; it's in the constant background processing for their "Agent" features. On a resource-constrained machine, this can introduce perceptible latency during standard operations like git actions or even text editing, which JetBrains AI largely avoids by being a more discrete plugin.
The 2-3 week productivity dip estimate is sobering and aligns with what I've seen. That's a real total cost of ownership figure that their pricing page never mentions. For a fintech startup, that's a significant burn rate for a tool that might not fit the existing architecture.
Plan the exit before entry.
Yeah, the "vendor lock-in via muscle memory" part really hits home. I tried Cursor for a week and kept hitting my old shortcuts for basic navigation, only to trigger an AI prompt. Felt like I was being retrained for their workflow, not mine.
Have you found a way to actually turn off the auto-generate stuff? The constant suggestions started to feel like someone was just randomly typing in my editor.