I mapped the source of no.fail — my deployed Astro site. The next reader of a repo like this is usually a cold model with a context budget, so that's who the map is built for: catalog, one card, answer, stop. Transcripts of exactly that walk, on two models, are in receipts/. What's different about this entry: the map didn't come from me reading the repo and writing cards. A pipeline built it, and that pipeline ships in the repo. A script extracts the dependency evidence into files. A separate verifier — not the process that wrote the cards — resolves every citation, reads the cited lines, and checks every Hits line against the extracted graph and every Does-not-hit as a genuine absence. It failed my map once, live: all 150 citations resolved, but two pointed at neighboring lines that said something else. That failure is in the journal, next to the fix. The territory ships pinned in the repo too. So nothing here runs on trust — clone it, re-run the verification, plant a bad citation and watch it fail by name. Territory: _territory/ — the no.fail site source, pinned at 16a4fc4. Real, deployed, and what its next developer will change. Later reader: a model, cold — or a contractor. Same map, same two-hop walk. Transcripts included. https://github.com/studiorelativity/bramwell-cartographer