What's this about?
CodeNib treats "what context does a coding agent need from this repo" as a data-serving problem — a multi-view system that assembles and serves the right slices of a codebase on demand rather than one-shot retrieval.
Why it matters
Context selection is the current ceiling on coding agents; framing it as an infrastructure/serving problem (56 upvotes) is a useful shift for anyone building agent tooling.