The Cross-File Reasoning Trap

Senior AI Engineer interview at Google DeepMind, and the interviewer asks:

“Your theorem-proving LLM is blind to a researcher’s custom project. You stuff the entire active file into the 128k prompt context. Local accuracy spikes, but it still fails on cross-file theorems. Why?”

Don’t say: “The context window isn’t big enough. Let’s switch to a 1M token model and dump the whole repository in, or build a standard dense vector RAG pipeline over the .lean files.”