2 approaches are emerging for giving ai agents structural understanding of a codebase.
1. LLM inferred. fast to build. but edges are labeled INFERRED because the model is just guessing. confidence scores include bc it knows it could be wrong
2. deterministic. built straight from the actual code. same input same output every time. no guessing
so my question is: why on earth is the inferred approach winning on github right now?? it seems clear to me which one wins. Does deterministic structure matter to you or is inference good enough.
Need to know based on what im building here https://github.com/ix-infrastructure/Ix
1. LLM inferred. fast to build. but edges are labeled INFERRED because the model is just guessing. confidence scores include bc it knows it could be wrong
2. deterministic. built straight from the actual code. same input same output every time. no guessing
so my question is: why on earth is the inferred approach winning on github right now?? it seems clear to me which one wins. Does deterministic structure matter to you or is inference good enough.
Need to know based on what im building here https://github.com/ix-infrastructure/Ix