Cursor local indexing VS competitors

Trying again, as the lovely moderators decided to delete my previous post because I was mentioning a competitor. Very open and mature mindset, well done.

So, without mentioning any competitor (it’s just one that’s all about being very effective with large codebases), I’ll post again here, as I see on forum others complain about how bad is Cursor indexing.

It’s getting more and more frustrating using Cursor on large codebases (about 1m lines of codes across a few thousand files). Although it says it indexed all the codebase, it fails to really look through it:

  1. often loads only partial files and needs to read them in multiple steps
  2. very often is unable to find things and starts using terminal with ‘find’ and ‘grep’

It just feels prehistoric (and it fails a lot), while competitors able to do a proper index and somehow always find what’s needed even in VERY large codebases.

Possibility of choosing model and all the other options in Cursor without a top-class local indexing makes it not very useful.

This is a question to understand if others find it frustrating that local indexing for Cursor is so bad on large codebases, and to the developers on why they don’t change their local indexing to something much better like some of their competitors (almost magic on codebases almost 1m lines of code and thousands of files!).

I myself do research in the code base using regular expressions and deliver it to it works well for me study regular expressions in the search and see magic happens

I do know them very well, I have 25+ years experience of sysadmin. That’s why I know they’re not nearly as good as a vector database where all the codebase is efficiently indexed.

And why results are always much poorer (and slower), that’s why I’m asking the reason Cursor has a very limited local codebase indexing capabilities compared to others.