Windsurf

What I like most is the seamless AI integration - especially how easy it is to get intelligent code suggestions and edits right within the editor. Features like single-file and multi-file AI suggesioins, model switching, and smart autocompletion feel like magic. The simplicity of setup, combined with powerful features, makes Windsurf feel like a next-generation coding tool.

It definitely solves a large problem in software development (and really, any soft of creative work): how to get started. The ability to start from a blank screen, type a prompt, and then watch code appear almost instantly, is pretty magical. Even if you use this code as a starting board, it definitely increases productivity right from the start.

The ability to select code and chat about it. The ability to change models on the fly and get slightly different results to questions that can provide more insight into problems. The ability for it to be able to watch the output in a limited manner and help troubleshoot setup or configuration issues.

1. Unreliable subscription system: Even with an active subscription showing 19 days remaining and auto-renewal enabled, the service suddenly stopped working for the entire team, showing no active subscription errors. 2. Delayed and unclear support response: Despite previous good experiences with support, this time I received no response for several days, even after creating a ticket. The delay added to the frustration, especially during active project work. 2. Team impact and trust: Because of the unresolved subscription issue, some team members lost trust and move to other tools. This experience damaged the positive impression I had built within my organization.

As I mentioned, the main issue is while it may grasp the particular code file you are working on, or the currently solution, it lacks awareness of other codebases in a large enterprise application that is common these days. It also suffers from a lot of the issues with AI these days: the desire to please no matter what, even if it means faking it, or pretending to understand. It also likes to go in loops, suggesting one thing to solve a problem, and then when that not working, suggesting something else, and if that doesn't work, loop around to the first.

The ability to do remote debugging even if it doesn't involve the AI. The ability to get AI feedback while debugging would be invaluable. Being able to show the AI what values would be expected at runtime would allow it to watch for issues leading up to that point. Better support for SQL integrations. It does a reasonable job but has to be watched closely as it seems to be guessing about what is happening in SQL. It forgets tables that need to be populated with data and doesn't seem to handle optimizations well. Join tables it kind of does ok, but aren't optimized well. It does seem to be good about reminding to do table updates when a new column is added so that existing deployments won't crash on missing columns.