Cursor used to be amazing. Now it's too aggressive with jumping into coding and ignores context. Windsurf just works better - shows what it's reading, follows instructions, and doesn't go off the rails. First AB test of Cursor and Windsurf in a real-world application. eric zakariasson Cursor Michael Truell

I'll keep testing tools and recording what works and doesn't work for me. I hope the Cursor team can steer the ship around. Or show me the way of what I'm doing wrong! TY

Boost, surf, and weave your way through The Art of Flight, an arcade game about flying multiple ships at the same time. With solo, local co-op, and a leaderboard, there are tons of ways to play. Wishlist on Steam today!

The incentive for all of these apps that charge a fixed dollar per month is to minimize how much is read into context, etc. to keep their costs down. Is @cursor_ai leading the way in succumbing to this? I'm pretty happy with @windsurf_ai - it definitely goes off the rails sometimes, but I'm pretty sure it's the model straying, not the IDE. Alternatives include using a bring-your-own-key IDE add-on and copy-pasting everything you want in context to a non-integrated LLM.

@cursor_ai @windsurf_ai Yeah I suspect that, but even if you use the MAX models with is pay by use it’s not working. In my video I’m using the MAX models

100% I noticed the same! But instead of @windsurf_ai , I compared to @CodebuffAI. I think Codebuff is the best coding agent today.

Windsurf is in top popularity growth compared to other similar tools

Been coding Ruby on Rails with Windsurf and its great. Have switched between Cursor and Windsurf often, but mostly on Windsurf because you feel like you're in a "state of flow" They do lack some features like putting rules into a folder. There's memories, but not exactly "exportable" to team members on the same project.

Yes 100%! State of flow can be achieved in Windsurf, and not in Cursor anymore

Have you tried

Not yet but heard good things

I watched your video and I definitely relate! I can still do small targeted tasks in Cursor, but idk. Yesterday I started testing Cline with Gemini, and there's something to like about that. It did as told, and when it didn't it got the correction right. Strange. I feel like Cursor trusted too much the "agentic" abilities know what they're doing, while in reality 3.7 does need to be kept in a tighter leash.

Yes 100%, 3.7 and MAX do not work great with their agent. 3.5 was great.