Curious if others are dealing with this: as more AI-generated code gets merged, the high-level structure of codebases seems to be quietly getting worse. Not syntax stuff, linters catch that fine. More like modules slowly getting tangled with things they shouldn’t know about, circular dependencies that sneak in, architecture that stops making sense over time.
I’ve been trying to solve this by treating it as a graph problem. Parse the imports, build a dependency graph, run it through a neural network pipeline that flags where the structure is breaking down.
It’s a side project, still rough around the edges. Would love to know if a terminal-based visualization of your dependency graph would actually be useful to you? Here’s the link to the product: https://gneiss-systems.vercel.app
답글 남기기