The source work behind the guides.

Twelve study notes from Maria Keet's ontology-engineering textbook, in book order, with the three consolidated chapter notes marked. Start here if you are entering the annex.

A tidy schema is not the goal. The model has to help an operator make decisions, hand off work, and run the business.

Reasoning is not garnish. It is part of how a business system proves the model still makes sense.

Formal models matter when they tell the team what the system can infer, explain, and afford to compute.

OWL is only one layer in a broader stack. If a company does not name the boundaries between data, validation, and reasoning, operators are left guessing.

Five study passes through Keet's methodology chapter converge on one lesson: consistency checks stop at the class layer, and the expensive failures live everywhere else.

A shared upper layer only creates leverage when the company states what it means, what it governs, and where it intentionally stops.

Business systems get dangerous when every contains-style relation is treated like structural part-of.

A database schema is not the business model, an AI label is not accepted knowledge, and a spreadsheet column is not a concept. Chapter 7 is about the review that stands between source and ontology.

Operators should not have to guess whether the ontology stores the truth, points to it, or only documents intended structure.

If Hadto wants ontology infrastructure that can travel across communities, multilingual support has to be treated as a governed operating layer—not an afterthought of translated strings.

Scores can help with review, but they do not change crisp semantics unless the platform says exactly what they mean.

Green checks are a layer, not a verdict. Chapter 11 is about the release discipline: readable trust, an inspectable research intake, and derivative surfaces that admit they are derivative.

Dental attachments are not generic uploads. They are labeled proof bundles tied to eligibility, authorization, claims, appeals, orthodontic continuation, and quality measures.

A waived requirement is not the same as finished work. A business needs to remember what changed, what remains, and who owns the next action.

A new payer code is not always a new business concept. Sometimes the right operating move is to inherit the old rule, not create fresh local lore.

A radiograph is not only an attachment and not only a code. The same image can be evidence, a covered benefit, a denied benefit, a bundled service, a documentation duty, and a frequency event.

The same tomosynthesis code family can be denied by default, covered by contract, mapped to a conventional L.E.A.T. benefit, or treated as capture-only non-benefit. The plan lane decides which rule is live.

Coverage is not the member balance. A practice still needs to know which benefit lane, buy-up, network rule, proof route, and billing prohibition produced the amount the member may owe.

A covered-code guide, clinical-review advisory, contract, prior-auth flag, and frequency counter are different rule sources. The operating system has to preserve which source is speaking.

A blank-looking review row is still a business fact. It tells the operator which question the document is answering, and which questions still have to be checked somewhere else.

A DentaQuest/Molina manual can run benefit tables, routing, authorization, and attachments while MDHHS remains the state authority. The system has to preserve both roles.

Healthy Kids Dental is a child benefit lane with its own administrators, proof duties, frequency groups, fee-schedule consequences, and member-billing guardrails.

Coverage tells the office one thing. The claim path tells it another: which form to use, which modifier is required, when attachments have to wait, and what a denial changes next.

A measure can name the gap. It cannot prove that a state, plan, provider network, or care team changed the workflow that produced the gap.

APIs may reduce portal and fax work, but the operator still needs to know which payer path exists, which requirements were discovered, where the request sits, and who owns the next action.

South Dakota's pre-orthodontic certification is not an HLD score, not a coverage approval, and not generic preventive history. It is a dental-home readiness record that has to travel into the orthodontic review path.

A family cannot use a covered dental benefit if the directory cannot tell them which provider, plan, location, specialty, language, and freshness promise can be trusted.

A Medicare Advantage dental answer is not just covered or not covered. The answer belongs to a contract, plan, segment, bid, version, and benefit section.

Iowa's adult dental ABM is not only a $1,000 limit. It is a paid-claim benefit accumulator with exclusions and emergent-service evidence duties that change scheduling, billing, and patient explanation.

The Dental Home is assigned through member, family, network, and quality facts. It also carries referral coordination, authorization timing, claims ownership, appeal routes, and billing duties.

D9995 and D9996 do not behave like standalone MaineCare dental benefits. They identify synchronous or asynchronous teledentistry on a claim while the underlying diagnostic service carries the payment question.

Alabama's D0999 line can pay only when it rides with covered dental services, while New York and Pennsylvania teledentistry rules show zero-dollar, retired, line-order, POS, and encounter-payment instructions.

NJ FamilyCare's 2026 grid binds D9995 to synchronous teledentistry, D0140, and specific member lanes. The source role matters as much as the code.

A network report can prove that the state is watching access. It does not prove that a patient can get the appointment, or that a code should be paid.

A covered sedation benefit is not enough. The operator needs the network standard, provider directory signal, appointment capacity, and source boundary in one inspectable record.

A Health Affairs study finds cutting Medicaid adult dental benefits causes 60-point uninsurance spikes that persist for years, while restoring benefits recovers only part of the loss. Dental coverage is not a reversible switch.

A state can expand adult dental coverage and still leave different members under different delivery routes. The benefit record needs the MCO, FFS, carve-out, network, and payment context.

G0330 and related state workarounds show a narrow operating lesson: the facility fee pays for the room and institutional use. It is not the dental visit.

Washington's Oral Health Connections pilot ended on December 31, 2023. The diabetes-specific periodontal-maintenance rule survived in narrower statewide form on January 1, 2024.

Our ontology research loop is already showing a useful pattern: home services looks like the fastest packaging opportunity, professional services looks like the deeper differentiation play, and dental remains the best place to keep learning.

Discovery-backed competency questions are still climbing, but maintenance proposals are being rejected by answer-path quality gates—exactly the kind of friction that protects the platform from scaling with weakly grounded ontology edits.

The newest research cycle did more than add competency questions. It showed the same escalation pattern recurring across dental, home services, professional services, and franchise operations—a strong sign that Hadto should look for a reusable cross-venture operating primitive.

A fully green ontology dashboard means we answered the questions already in scope. It does not mean we have finished learning what future owner-operators need from the platform.

A research loop becomes an asset only when every claimed discovery can be traced back to source material and review.

Discovery is only intake. Real learning happens when evidence survives review and changes the shared method the next operator inherits.

A research loop starts to matter commercially when it can turn model pressure into a ranked offer, a buyer hypothesis, a discovery script, and a proof surface another operator can run.

A premium service line stops depending on private judgment when the business names where work is blocked, who owns the review, and what requires manual follow-up.

The useful proof after a repair is not that the machinery ran. It is that the next operator can see the decision, the remaining risk, and the path to judgment.

A green dashboard can protect pride. A teachable business needs something harder: a system that can admit a new gap, keep the evidence visible, and show the next operator where judgment belongs.

A quiet queue can mean safety, failure, rejection, or blindness. Future owner/operators need systems that preserve the reason for absence.

Karpathy's autoresearch loop is useful because it keeps score. Ontology work needs its own score: less uncertainty without breaking the commitments the business already runs on.

The first OntoMoBench proof is a shipped benchmark contract. The next proof is a deterministic scorer that checks syntax, runtime behavior, transition conformance, and invariants against business traces.

OntoGPT did not teach us to let a model invent the ontology. It taught us to bind suggestions to schema, source text, competency questions, and scores that catch unsupported facts.

Hadto's internal proof is a documented overlay contract, JSON schema, deterministic scorer, and fixture-backed tests for three operator surfaces.

Mo Gawdat's AI thesis is easy to caricature. The more useful read is that he is testing a harder claim: the builder's ethics matter most when the system is already leaving the lab.

The Tahoe signal is not one clean causal story. It is an operator-facing case about tax risk, luxury property, data-center load, and capacity limits starting to rhyme.