Findings

Five publishable headlines from the v6 catalog — 912 records × 68 columns across 34 sections, 528 typed edges. Each card cites the analytical view that produced it; full derivations live in analysis.md.

  1. 1 of 100 priority-cohort products

    Only LangChain's SemanticCache ships a true semantic-cache layer in the v6 priority cohort. The cleanest product gap in the catalog: a vendor-shipped Anthropic / OpenAI semantic cache with cross-model similarity would have no direct competitor.

    Read the full finding → See the cost-economics matrix analysis.md §25.2 · commit f2b95c1
  2. 833 of 912 products · only 2 scores on a neutral leaderboard

    The product × benchmark matrix is 119 × 25 with 169 filled cells — the story is the empty cells. Of the 169 filled, the integrity split is 111 peer-reviewed / 2 independently-verified / 52 vendor-claimed / 4 disputed. Almost every claim is a paper or a vendor blog.

    Read the full finding → See the product × benchmark matrix analysis.md §24.2 · commit ea70f89 (T1-5)
  3. Inbound runtime-deps: Claude 62 · GPT 52 · MCP spec 34 · Gemini 22 · Qwen 16

    From the T2-1 runtime-dependency graph (212 edges of 'X depends on Y at runtime'). A protocol specification ranking #3 — ahead of every foundation model except Claude and GPT — is the v6 finding that wasn't visible in citation-graph hubs. The MCP spec is becoming a substrate, not just a protocol.

    Read the full finding → See the runtime-dependency graph analysis.md §23 · commit ddb26c7 (T2-1)
  4. 0 inbound edges · 0.71 normalised betweenness — highest non-trivial in the graph

    Sits on the shortest path between the MCP-spec substrate cluster and the Zep / Graphiti citation-anchor pair. A high-betweenness / low-inbound node is the structural definition of an under-acknowledged-but-load-bearing connector. Top-5 bridge-surprise: Graphiti MCP Server, MAGMA, Memformers, MemEvolve, RGMem.

    Read the full finding → See the bridge-surprises callout analysis.md §26.2 · v6 centrality view
  5. 108 of 140 FM-dependent rows depend on OpenAI / Anthropic / Google

    Of the 140 rows (~16% of non-FM rows) that name a foundation model, 77.1% sit on the OpenAI / Anthropic / Google single-vendor-risk tier. The long tail (Qwen 16, Mistral 12, Cohere 8, DeepSeek 6, Jamba 3, Grok 2, Nova 2) is sparse. The substrate-risk map is concentrated on three labs.

    Read the full finding → See the substrate-dependency risk panel analysis.md §19.5 + §21.3 · v5/v6 FM substrate mining