The headline is type User = z.infer<typeof UserSchema>: the schema becomes the source of truth, and the static type follows automatically. Real Zod's z.infer is a conditional type that walks the schema's runtime shape and reconstructs the equivalent TypeScript type, so adding email: z.string() to the schema instantly makes User.email available downstream without a separate type edit. The runtime stubs here keep the validator running offline; in your project, remove the stubs and import from zod directly. This pattern is the foundation of every typed boundary in modern TS apps (form validation, API request bodies, env var parsing).

Schemas with transforms have a different input type than output type: z.string().transform((s) => Number(s)) accepts string but produces number. Real Zod ships z.input<typeof S> and z.output<typeof S> (and z.infer aliases z.output) so callers can pick the right shape per use case. The handwritten InputOf / OutputOf here are the same conditional infer pattern, simplified for the snippet. Use the input type at the boundary (request body parsing) and the output type after parsing succeeds (downstream business logic).

Wrapping schema.parse in a Result<T> gives callers a discriminated union to switch on instead of a thrown exception, which is friendlier for HTTP handlers, form validation, and any boundary that wants to map errors to status codes. Real Zod ships schema.safeParse(value) with the same shape ({ success: true, data } or { success: false, error }), so callers can move the same Result pattern in and out without rewrites. The generic <O> flows from the schema through the wrapper so the success branch is fully typed without any cast at the call site.