Windsurf, Cursor, Claude Code – I do believe that it’s more important than I’d like

I have used all three during the year in search of the tool that would make my codebase manageable but ended up with an unpleasant conclusion – the problem wasn't in the tool. It was in the lack of tests in my codebase.

Any agent, no matter what brand it is, becomes significantly better or worse depending on the confidence in its ability to validate the changes made by itself. And since my codebase didn't contain any tests six months ago, all agents seemed to be unreliable in the same way, having nothing to verify themselves.

I started using a proper test suite last month prior to changing the tool once more and noticed that the reliability difference turned out to be greater than ever. Does anyone else find the "tool debate" just a smoke screen?