Array.prototype.at(index) accepts negative indexes that count from the end, which is the cleaner spelling of the classic array[array.length - 1]. It is supported on every evergreen runtime (Node 16.6+, all modern browsers). The savings get bigger when the array is the result of a computed expression: getRows().at(-1) runs the call once, while getRows()[getRows().length - 1] runs it twice and breaks if the function is non-idempotent. It also reads as plain English, which matters more than micro-optimisation.

Both array.at(-1) and array[array.length - 1] return undefined for an empty array, which is fine until you destructure or call a method on the result. A tiny last(array, fallback) helper makes the empty case explicit at the call site, and TypeScript users get a clean T | F return type for free. This is one of those three-line utilities worth committing to a utils/array.ts rather than re-deriving in every component. Keep fallback as a value, not a thunk, unless you genuinely need lazy evaluation.

The bug that hides in array.pop() is that it mutates the array. If you only meant to read the last item, at(-1) is the right choice and your function stays pure. Reach for pop() only when stack semantics are actually wanted (undo history, depth-first search, expression evaluation). When the array is Object.freezed or an immutable proxy, pop() will throw or silently fail, while at(-1) always works. The same distinction applies to shift() versus at(0) for the front of the array.