Appendix · IV·Glossary

The vocabulary,
without the code.

Every Buildable audit speaks two languages at once — the engineer’s and the operator’s. This is the operator’s dictionary. No code. No condescension. Just what each term means when it shows up in a board deck.

AI-Native: Software where AI is a load-bearing part of the experience, not a sidebar feature. AI-native products are designed for prompts, embeddings, and agents — not retrofitted with a chatbot.
API: A contract that lets one piece of software call another. When Stripe ‘has an API,’ it means your app can charge a card by sending a structured request, instead of a human filling out a form. APIs are the plumbing of every modern integration.
Auth: Two questions in one word. Authentication: who are you? (login, password, Google). Authorization: what are you allowed to do? (admin vs. viewer). Most outages and breaches happen at the seam between the two.
Buildability Index: A 0–100 index measuring how standard the product's surface is — recognizable UI patterns, commodity data model, shallow integrations, simple auth, public APIs, coherent visual identity. Tool-agnostic: it reads the *shape* of the stack, not whether any specific builder can ship it. High: a competent operator could reconstruct the surface in any modern AI-native builder. Low: real engineering depth (custom engines, exotic data, deep enterprise integrations).
Churn: The percentage of customers who cancel in a given period. A buildable surface lowers the cost of being replaced. A real moat lowers churn regardless.
CRUD: The four operations any database-backed product needs: add a record, view it, change it, delete it. When an audit says ‘the product is mostly CRUD,’ it means the visible workflow is essentially a styled form over a list — buildable in days, not quarters.
Also called · Forms over data
Edge Function: A small piece of server logic deployed to data centers around the world, so the response reaches the user in milliseconds. Replaces what used to require a dedicated backend team and a fleet of servers.
Frontend / Backend: Frontend is the browser experience (buttons, layout, animation). Backend is everything the user never sees (database, billing logic, integrations). A ‘full-stack’ change touches both.
Hallucination: An LLM producing plausible-sounding output that is wrong. The risk that turns ‘AI can do it’ into ‘AI can do it, with a human in the loop.’
LLM: The class of AI behind ChatGPT, Claude, Gemini. Trained on enormous amounts of text, it predicts the next word well enough to draft, summarize, classify, and now — increasingly — write working code.
Lovability Fit: A 0–100 score measuring fit with Lovable's native primitives — component model, prompt one-shot efficiency, Supabase integration depth, TanStack routing & auth alignment, generation edge-case profile, and iteration cost from seed to production. This is the Lovable-specific score; Buildability measures the surface's commodity shape regardless of tool.
Migration: A script that takes the database from one shape to the next, recorded so every environment (developer laptop, staging, production) ends up identical. Migrations are how teams change the schema without losing data or sleep.
Moat: Data accumulated over years, integrations the customer cannot easily reproduce, regulatory approvals, distribution advantages, switching costs. The reason a buildable surface does not equal a replaceable company.
Moat Density: A 0–100 score measuring data network effects, integrations, regulatory posture, distribution, and switching costs — the parts that survive a rebuild. A product can be highly buildable and highly defended at the same time.
Multi-tenant: All customers run on the same codebase and database, separated by access rules (see RLS). Cheaper to operate, harder to get security right. The default for modern SaaS.
MVP: The smallest version of a product that delivers the core promise and can be tested with real users. Not a prototype, not a demo — something a paying customer could actually use, however narrowly.
Network effect: Each new user makes the product more valuable to every other user (Slack, Figma multiplayer, marketplaces). The most durable kind of moat — one that compounds, and one that a rebuild cannot ship on day one.
Prompt: In an AI-native build, the prompt is the new specification document. A good prompt produces a working MVP. A bad prompt produces a demo that breaks on the second click.
RLS: A database rule that says ‘this user can only see their own rows.’ It moves access control from the application code into the database itself, so a bug in the UI cannot leak another customer’s data. The standard for multi-tenant SaaS.
SaaS: Software you rent, not own. Delivered over the web, billed monthly or annually. The dominant business model of the last twenty years — and the model Buildable audits exist to read.
Schema: A formal definition of what fields a record has, what types they are, and how tables relate. The schema is the skeleton of the product — change it carelessly and everything downstream breaks.
Seed Prompt: The single, structured prompt that a Buildable audit emits. Paste it into Lovable and you get a working first version of the audited surface. The seed is the audit’s deliverable.
SSR / SSG: How a page is built before it reaches the browser. SSR builds it fresh for each request (good for personalized dashboards). SSG builds it once at deploy time (good for marketing pages and SEO). The choice affects speed, cost, and how Google sees you.
Stack: Frontend framework, backend runtime, database, hosting, auth provider. Saying ‘React + Supabase + Vercel’ is a stack. The stack determines hiring, cost, speed of iteration, and what you can rebuild in a weekend.
Stack of record: When a tool becomes the source of truth — the place finance reconciles, legal references, ops opens every morning — it is the stack of record. Hard to displace, even if the surface is buildable in a weekend.
Surface: Pages, forms, tables, charts, settings. Surface is the part of a product an audit can decompose and rebuild. The rest is moat.
Switching cost: Migration work, retraining, lost integrations, contract penalties, political capital. Often higher than the price of the new tool — which is why buildable does not mean replaceable.
Translation: The new craft Buildable measures. Anyone can reproduce a screen. Translation is decomposing a product into the smallest prompt that rebuilds its essential workflow — and naming what cannot be rebuilt.
Webhook: Instead of your app asking ‘did anything happen?’ every minute, the other service calls you the moment something happens. Stripe webhooks tell you a payment cleared. Slack webhooks post a message. They make integrations real-time without polling.

Missing a term? · Open an issue on the framework repository — the glossary ships with the audits.

The vocabulary,without the code.

The vocabulary,
without the code.