Over the past few weeks, I’ve noticed a clear boost in my productivity. I’ve been using Claude Sonnet 3.5/7 but now also Gemini Pro 2.5, and I’ve been genuinely impressed. These models perform reliably: they understand context well, handle multi-file edits gracefully, and often go above and beyond—catching bugs or edge cases without being explicitly asked. Sure, there are still some frustrating cases of errors and mistakes, but they’re becoming fewer and fewer, and certainly doesn’t prevent huge steps toward the goal of project completion.
What’s important to highlight is that the IDE matters. These same models, when run in anoter (commercial) IDE, behaved poorly in my latest test—creating duplicate files instead of editing current ones or mismanaging context. It really showed me that it’s not just about having a smart model; it’s about how it’s integrated. Cursor has clearly nailed this part.
So, a big thanks to the Cursor team. When something works this well, it deserves to be acknowledged.