
graphify token reduction benchmark
──────────────────────────────────────────────────
  Corpus:          74,500 words → ~99,333 tokens (naive)
  Graph:           1,490 nodes, 2,526 edges
  Avg query cost:  ~1,464 tokens
  Reduction:       67.9x fewer tokens per query

  Per question:
    [124.8x] how does authentication work
    [44.2x] what is the main entry point
    [38.2x] how are errors handled
    [111.7x] what connects the data layer to the api
    [125.7x] what are the core abstractions

