The fastest way for a calculator to wander into regulated territory is to invent an insurance number. The models that touch insurance here share one discipline: real quotes in, decision analysis out, and no premium ever invented.
Insurance is where calculator scope discipline matters most, because the regulatory texture changes underfoot. Product illustrations — the projections of policy values an insurer or producer shows a consumer — are regulated at the state level, shaped by NAIC model regulations that govern what may be depicted, on what assumptions, with what signatures and disclosures; life insurance illustrations have their own model regulation, and other lines carry their own rules. A general-purpose financial calculator that starts generating policy-value projections or estimating premiums has drifted toward a regime built for licensed activity. This piece maps how Worthune's insurance-adjacent models are scoped to stay clear of that drift, and what an integrator in the insurance world still owns.
The design rule: quotes are inputs, never outputs
The long-term-care model is the clearest case. It does not price a policy, estimate a premium, or project policy values — it takes a real quote as input (the monthly premium, the daily benefit, and the benefit period, straight from a quote sheet a licensed producer provided) and computes the decision arithmetic around it: the compare-against-self-funding analysis a household needs after the quote exists. Its comparison verdict — the spec's isInsuranceWorthIt boolean — is that break-even arithmetic under stated assumptions, not a suitability judgment; suitability weighs a person's circumstances no cost comparison sees. The estate-planning model applies the same rule to its insurance-shaped edges: user-supplied figures, applied as given. Nothing in the catalog underwrites, rates, or illustrates. The what-we-do-not-model piece (/writing/what-worthune-does-not-model) states the general principle — a published spec cannot honestly describe a number a third party sets privately — and insurance is where that principle has the sharpest regulatory shadow, because the number a tool might be tempted to invent is precisely the number whose presentation state law regulates.
| Activity | Which side of the line | Who owns it |
|---|---|---|
| Quoting or estimating a premium | Illustration territory | The insurer and licensed producer |
| Projecting policy values or dividends | Illustration territory | The insurer, under state illustration rules |
| Comparing a real quote against self-funding | Planning analysis | The catalog's LTC model computes it |
| Testing estate liquidity with user-supplied figures | Planning analysis | The estate model computes it |
| Deciding coverage suitability for a person | Advice territory | A licensed professional, never a calculator |
Why the boundary is architectural, not editorial
The discipline would be fragile if it lived in marketing copy. It lives in the contracts instead. The LTC model's premium is a required input with a published domain — there is no code path that produces a premium, so no integration, prompt, or UI misuse can extract one. The specs' exclusions name what is out of scope, the envelope's assumptions restate the model's frame, and the platform disclaimer rides on every response. An integrator who wants policy projections cannot coax them out of these models; they must go get them from an insurer's illustration system, which is the correct outcome. Scope enforced by input surface is the same mechanism that keeps the catalog honest everywhere else — here it doubles as regulatory drift protection.
The household-side integrations
For non-insurer products — planners, coaches, benefits platforms — the discipline translates into three habits. Collect real quotes: the UI should ask for the premium from an actual quote sheet, and say so, because "estimate your premium" fields are invitations to invent the regulated number upstream of the model. Frame outputs as decision analysis: the LTC model's answer is about the user's alternatives given a quote, not about what any policy will do. And route product questions outward: when a user asks what a policy would cost or pay, the honest response is a referral to a licensed source, in the same scope-refusal pattern the assistant guides develop (/writing/refusing-beats-guessing).
- Never present a model output as a premium, policy value, or illustration.
- Collect real quoted figures, labeled as such, for every insurance-shaped input.
- Keep illustration surfaces and planning surfaces visually and verbally distinct.
- Route pricing and suitability questions to licensed channels.
- Put this boundary in your compliance review explicitly — it is a state-law surface, and this piece is not legal advice.
The pattern, generalized
Every regulated vertical eventually poses the same question to a computation platform: will you generate the number our regulator cares about most? The durable answer is the one this pillar keeps arriving at — no, and here is the architecture that makes the no credible. Real inputs, published scopes, rejected out-of-domain requests, and a paper trail for everything computed. Insurance is simply the vertical where the alternative — a confident invented number wearing a calculator's authority — has the clearest name in law, and the clearest reason to be refused.
Sources
- [1] NAIC, Life Insurance Illustrations Model Regulation (Model #582). https://content.naic.org/model-laws
- [2] Worthune writing: What Worthune Deliberately Does Not Model. https://worthune.com/writing/what-worthune-does-not-model
- [3] Worthune long-term-care model documentation. https://worthune.com/docs/models/long-term-care