Another day, another "AI-powered" tool that can't handle a real codebase.
My indexing gets to 5% and just sits there. No errors. No progress. CPU usage drops to nothing. Tried it on three different repos of varying sizes.
Common factors:
* Using the default configuration.
* Happens with both local and GitHub repos.
* Restarting the app or re-adding the project does nothing.
Checked the usual suspects:
* File permissions are fine.
* No weird symlink structures.
* Exclude patterns are default.
Is this just a hard size limit they aren't disclosing? Or is the indexing engine fundamentally broken for anything beyond a toy example?
- Prove it.
Your "prove it" challenge misses the core issue: you haven't provided the single most critical data point for any real evaluation - the actual size and composition of those repos.
A 5% stall could be hitting a pathological file - a 2GB minified JS blob, a corrupted binary, or a circular symlink your check missed. Without knowing what's in the queue at that moment, you're just guessing.
The real problem with these AI coding tools is their opaque benchmarking. They're tested on clean, curated datasets (think the The Stack deduplicated), then break on the messy reality of a working codebase with build artifacts, decades of legacy code, and vendor SDKs.
What does "varying sizes" mean? 10k files? 100k? Show the actual file tree output at the point of failure, or run a simple profiling script to see what it's choking on. Otherwise, we're all just trading anecdotes.