This will change things in both files, for example it will change a function and if that also needs to be changed it will change the place where the function is called.
If you’re working on a small app all your code in LLM will be across the entire app, but if you’re working on a large app – for example Recursy has around 80,000 lines of code – technically it can all fit in Gemini 3.5, although that’s a bit of a stretch.
But you also have lots of tools for choosing which files to include and getting an idea of what’s going on. Sometimes this results in things not working or crashing, but it’s really nice to be able to send that information back to the LLM so it can figure out what file needs to go and pull in to fix it.
Honestly, I could go on with this entire process for a long time, but this problem, if not 100% solved, is very close to it. It’s extremely rare that it breaks something in that sense, and when it does it’s quite easy to fix – it’s good at pulling not only files but also error messages.
I’ll go ahead and tell you about a process I use when working on really big projects, which is to take the entire project and create a prompt that contains all the code for the entire thing, and drop that into Gemini. And I ask Gemini to analyze the problem I’m talking about, give me the full path to all the files that are involved, even if peripherally, as well as give some analysis on how it can be fixed. But don’t write any code.
Then I take what that gives me to paste the entire output into a special tool in Recursy and it will create a new prompt that includes all the file paths seen in that output, i.e. it will take all the codes for all those files, or if I want to set it to only get signatures or documents I can do that. And then I’ll use that as the start of a new thread where it’s more focused and doesn’t have to be 100,000 lines of code, but maybe more than 10,000 that is relevant to the specific problem. And I think when you do that it becomes a little bit smarter because it’s not devoting a lot of its thought process to things that are completely irrelevant.
So there are a lot of tricks like this, but they only really come in handy if you’re working on larger apps, if you’re working on smaller apps, this problem is basically non-existent. If you watch the video where I tell it to take a 3D model and turn it into one of those fun explosion effects or whatever, you’ll see that it affects multiple files and it changes a whole bunch of methods in different files at the same time. So it knows what it’s doing if you provide it the context correctly and it doesn’t take a lot of effort on the part of the user, most of the tedium is handled by the app itself.
<a href