Ask HN: Thoughts on combining static analysis and AI for system understanding?
Working across multiple repos means constantly losing context. Most tools either search docs or try to feed entire codebases to LLMs (which fails).
I'm building something that uses static analysis to map relationships first, then gives that structured context to AI. For example: "what breaks if I change this auth function?" or "what code do I need to touch to add billing?" and it traces dependencies and business logic across all your repos.
Always wanted to leave the developer community better than I found it. This is a problem I've personally felt for years: context gathering. Good approach or am I missing something obvious?
Early version: https://www.omnispex.dev/