PlaybookA close reading
The product table says "Intelligence". The opening sentence says "AI framework".
The product table names six of eleven products "Intelligence" and never says "AI". The opening definition says "AI framework" and stacks five abstractions before a reader can point at anything.
Word counts in the playbook
Whole-word matches in KG's draft as captured on 2026-09-02.
"AI" appears eight times and always as the category label. "Intelligence" appears forty-five times and names the products. "Generative" and "decision" do not appear at all.
Source: the Google Doc · verbatim capture in the vault
What the graph of the document shows
We loaded the playbook text into InfraNodus. Eight clusters, modularity 0.45 (how cleanly the text splits into topics), and one word doing most of the connecting.
Betweenness centrality: how much of the document's meaning passes through the word. "Sovereign" and "agnostic" connect to nothing outside the opening paragraph.
01Finding
The opener is the weakest sentence in the document
"Model-agnostic sovereign AI framework with its own native intelligence layer that protects and compounds institutional judgment, alpha, and prioritizes security and IP protection." A reseller cannot repeat that on a call. A rewrite that a partner can say out loud: "Shur reads the public record of a market, scores it, and keeps what it finds. The customer keeps its own keys, data and memory."
02Finding
Security and judgment are cut off from the catalog
InfraNodus finds three gaps and every one has the security, IP and judgment cluster on one side. No product row says what protects the customer's intellectual property. What the graph points to: the COMPOUND layer and the security claim are one sale. Findings that stay the customer's own are the reason the memory is worth paying for.
03Finding
"Sovereign" now means something else
Since late 2023 "sovereign AI" is the phrase chip vendors and then governments use for national compute. A buyer who reads the trade press will hear that. Say the concrete thing: the customer keeps custody of keys, data and memory.
04Finding
Three meters, one worked example
A customer can pay a per-run price, a per-million-token Shur compute price, and a 15% model-management fee. The example in section 6 shows only the first. A reseller needs one worked example with all three on it, or resellers will explain the run economics wrong.
05Finding
No line for the buyer who is tired of AI
The document never says what kind of system this is. A reseller who hears "we are not buying another chatbot" has nothing to answer with. The product table gives the answer already: scored, evidence-backed, remembered. Give the partner one line that says which parts are drafted by a model and which parts are computed.
06Also
Small things
Three product names in one document: Shur, ShurAI (section 9) and the framework itself. "Executive operating surface" reads as a countertop to a buyer; say "dashboard". "Custom Intelligence System $15,000+" needs a minimum price and a scoping rule. No customer profile anywhere: who buys the first $25,000?
LineageTwo research lines
The product comes from the older line. It uses the newer one.
The public argument about AI is about text generators, from the 2017 transformer paper to the 2022 launch of ChatGPT. Shur's structure comes from decision support, cybernetics, semantic networks, knowledge graphs and network science, a line that has run since 1948.
Reasoning, decision support and structure1948 to 2026 · full span
Generative language models2016 to 2026 · the last ten years, zoomed
The lower line covers ten years; the upper covers seventy-eight. Shur is built on the upper line and uses the lower one for reading and drafting.
Where the terms come from
Decision support system: Gorry and Scott Morton, Sloan Management Review, 1971. Intelligence amplification: Ashby, 1956. Augmenting human intellect: Engelbart, Stanford Research Institute, 1962. Dynamic knowledge repository: Engelbart's Bootstrap work, 1990s, and the name our own DKR carries. Semantic network: Quillian, 1968. Cognitive architecture: Soar, Laird, Newell and Rosenbloom, 1983. Decision intelligence: Lorien Pratt, about 2008; Kozyrkov at Google, 2018; then a Gartner category.
What we can claim honestly
Graph structure, betweenness and modularity are computed with deterministic algorithms, so "algorithmic" and "network science" are true. Ontology as ground truth and claim-to-evidence-to-source chains are built and running, so "evidence-traced" is true. Formal causal inference, in Pearl's sense, is not something we run today, so "causal reasoning" is a claim to earn before we print it.
What the backlash is about
Fabricated citations, confident wrong answers, and text that reads like every other text. Each maps to a thing we already do differently: sources are cited by chain, scores are computed and can be traced, and the voice rules keep the prose from sounding machine-made. The naming should point at those three, and the demo should show them.
Names52 in seven families
Seven families. Filter by family or by who you are talking to.
Each card gives the name, a sentence you could say with it, where the term comes from, whether we can claim it today, how far it stands from the word "AI", whether a buyer already has a budget line for it, and the readers it fits. Star the ones worth taking on the next call.
Distance from "AI" is a judgment call on a five-step scale: five means a reader would never file it under chatbots. Budget line says whether a buyer already has a place in the budget for something with that name. Honest says whether the product does the thing today.
The same list as a table: category-name-options, in the vault
ConceptsEight, by reader
Eight things the product does, and how to say each one to six readers.
These are the concepts every name above is trying to point at. Pick a reader tab to see the sentence for that person, and use the three forms underneath when you need a headline, a slide caption or a spoken line.
Start hereSix to test first
Six names worth testing first, and the order to test them in.
Each has a budget line or a lineage behind it, each is true of the product today, and none needs the word "AI" to make sense. Test them on the next three calls before choosing.
01Category
Decision intelligence
A buyer already has a name and a budget for it. It says what the output is for. The risk is a crowded field, so pair it with "structural" when the audience is technical.
02Lineage
Decision support system
The oldest honest description. It dates the idea to 1971 and reads as sober. Use it with CEOs and boards who remember what the phrase means.
03Ours already
Structural intelligence
Already in the deck and the public tagline. Keeps continuity with everything written since March 2026. Needs one plain sentence after it every time.
04Backlash answer
Intelligence that shows its sources
The one line that answers the "not another chatbot" objection directly, because the evidence chain is built and can be demonstrated live.
05KG's line, minus one word
Enterprise intelligence infrastructure
The playbook already ends on "enterprise AI infrastructure". Drop "AI" and it describes the access-build-run model exactly.
06KG's fourth layer
Compounding intelligence
The COMPOUND layer is the strongest idea in the playbook. Use it as the outcome name under whichever category name wins.
What to change in the playbook this week
| Where | Now | Change to |
|---|---|---|
| Section 1, first sentence | "model-agnostic sovereign AI framework with its own native intelligence layer..." | "Shur reads the public record of a market and a company, scores it, and keeps what it finds. The customer keeps its own keys, data and memory, and can bring any model." |
| Section 1 | No line for the skeptic | Add: "Generative models do the reading and the drafting. The structure and the scores are computed by algorithms you can rerun, and the memory is ours." |
| Section 3, catalog | No row says what protects IP | One sentence under the table: every run's findings and memory belong to the customer and never train a vendor model. |
| Section 6, example | One meter shown | A second worked example with a run, Shur compute tokens and the management fee on one invoice. |
| Section 8, "approved Shur story" | "sovereign framework, model agnostic..." | "customer keeps custody, any model, deployable systems, metered runs, memory that compounds" |
| Section 9, row 5 | "ShurAI" | One product name throughout |
| Closing line | "enterprise AI infrastructure" | "enterprise intelligence infrastructure" |
The open question
Which of the six survives three real calls. The board on this page is where the answers go: star the names, leave the notes, and copy the shortlist into the next working session.
SourcesWhere the data comes from
Every date, count and attribution, with a link where one exists.
- Kristine (KG) Hagedorn, Shur Authorized Seller Playbook, Google Doc, captured 2026-09-02. Verbatim capturevault
- InfraNodus graph shuriq-category-naming-2026-09, built 2026-09-02 from the playbook text. 150 nodes, 761 edges, 8 clusters, modularity 0.45. Graph numbersvault
- The facts behind this page, numbered with sources: concept bundle and typed statementsvault
- A. M. Turing, "Intelligent Machinery", 1948 report, National Physical Laboratory.print
- W. Ross Ashby, An Introduction to Cybernetics, 1956, on intelligence amplification.
- J. C. R. Licklider, "Man-Computer Symbiosis", IRE Transactions on Human Factors in Electronics, 1960.
- Douglas Engelbart, "Augmenting Human Intellect: A Conceptual Framework", SRI, 1962; Dynamic Knowledge Repository, Bootstrap Institute, 1990s.
- M. Ross Quillian, "Semantic Memory", 1968, in Semantic Information Processing, MIT Press.print
- G. Anthony Gorry and Michael S. Scott Morton, "A Framework for Management Information Systems", Sloan Management Review, 1971; Scott Morton, Management Decision Systems, Harvard Business School, 1971.
- Stafford Beer, Brain of the Firm, 1972, the Viable System Model.print
- Gordon Pask, Conversation Theory, 1976.print
- John Laird, Allen Newell, Paul Rosenbloom, Soar, 1983 onward.
- Thomas Gruber, "A Translation Approach to Portable Ontology Specifications", Knowledge Acquisition, 1993.
- Duncan Watts and Steven Strogatz, "Collective Dynamics of Small-World Networks", Nature, 1998.
- Judea Pearl, Causality, 2000; with Dana Mackenzie, The Book of Why, 2018.
- Tim Berners-Lee, James Hendler, Ora Lassila, "The Semantic Web", Scientific American, 2001.
- Dmitry Paranyushkin, "Identifying the Pathways for Meaning Circulation using Text Network Analysis", Nodus Labs, 2011.
- Ashish Vaswani and others, "Attention Is All You Need", 2017. OpenAI, GPT-2, 2019; GPT-3, 2020; ChatGPT launch, 2022-11-30.
- Lorien Pratt, Quantellia, decision intelligence, about 2008; Cassie Kozyrkov, decision intelligence at Google, 2018; Gartner, decision intelligence and continuous intelligence as named categories, 2019 onward.
- Nvidia, "sovereign AI" from its November 2023 earnings call onward; national compute policy in several countries from 2024.
- Shur description corpus, 118 positioning statements, 2026-05-06. Corpusvault
Links marked vault open in Obsidian on a machine that has the totem-terminal vault. Items marked print have no stable public copy.