By defining Omit as its own conditional type, we can get a slightly nicer display, though it will introduce duplication of code between the two helpers.

Actually it's the mapped type - you'd want to inline the Pick, not the Exclude. So:

type Omit<T, U> = {[K in Exclude<keyof T, U>]: T[K]};

❤️

Yeah, I figured it out as I made the change.

Looks like this is getting reverted because of #31190.

Maaaaybe something like this?

// default type parameter trickery
type Omit<T, K extends keyof any, X extends keyof T = Exclude<keyof T, K>> = {
  [P in X]: T[X]
};

interface Test {
  required: string;
  optional?: string;
  readonly viewonly: string;
}

declare const x: Omit<Test, "required" | "optional" | "notPresent" >;
x.viewonly = "okay"; // error!
// Cannot assign to 'viewonly' because it is a read-only property.

// typeof x is Omit<Test, "required" | "optional" | "notPresent", "viewonly">>

Pro: The type parameter X maintains the homomorphicitinessitude that you need in the mapped type.
Pro: The name Omit will appear instead of Pick and Exclude in IntelliSense

Con: Default type parameters are not meant for this, who knows what could happen
Con: IntelliSense will show the value of X which could be confusing.

Or what about

// conditional type inference antics
type Omit<T, K extends keyof any> = Pick<T, Exclude<keyof T, K>> extends infer U
  ? { [P in keyof U]: U[P] }
  : never;

interface Test {
    required: string;
    optional?: string;
    readonly viewonly: string;
}

declare const x: Omit<Test, "required" | "optional" | "notPresent">;
x.viewonly = "okay"; // error!
// Cannot assign to 'viewonly' because it is a read-only property.

// typeof x is { readonly viewonly: string; }

Pro: Still homomomomomorphic
Pro: The name Omit appears instead of Pick, and in the concrete cases the actual distinct mapped type shows up.

Con: Conditional type inference isn't exactly meant for this either, who knows what can happen

The following type

type Omit<T, ExcludedKeys extends keyof any> = {
    [K in keyof T as Exclude<K, ExcludedKeys>]: T[K]
};

now maintains modifiers thanks to @ahejlsberg's pull request at #40633.

I think this change will be contingent on performance tests though.

Also, @weswigham may want to weigh in since he has a PR out at #37608, but I think it would be undesirable to add a 3rd type parameter just to use a default type argument.

For TS 4.0 and older, you can use @Jack-Works’s type-challenges/type-challenges#141.