Ordinals generalize natural numbers to capture both size and order structure, extending the sequence 0, 1, 2, ..., ω, ω+1, ω+2, ... Ordinals are well-ordered sets providing the foundation for transfinite recursion and induction, allowing finite techniques to be applied to infinite structures.
Study small ordinals: 0 = ∅, 1 = {0}, 2 = {0,1}, 3 = {0,1,2}. Visualize ω as the 'first infinite ordinal' followed by ω+1, ω+2. Practice comparing ordinals using the membership-based ordering.
To understand ordinals, start from a question your work with set membership already makes natural: can we build *numbers* purely out of sets? Von Neumann's answer is a beautiful recursive construction. Define zero as the empty set: 0 = ∅. Then define each successive number as the set of all numbers that came before it: 1 = {0} = {∅}, 2 = {0, 1} = {∅, {∅}}, 3 = {0, 1, 2}, and so on. Every natural number is a finite set whose members are exactly its predecessors. Crucially, the ordering between ordinals is just set membership — n < m if and only if n ∈ m. You don't need a separate definition of "less than"; it falls out of the set structure itself.
What makes ordinals more than a clever encoding of the natural numbers is that this construction doesn't have to stop. After all the finite ordinals 0, 1, 2, 3, ..., there is nothing preventing us from forming their union: ω = {0, 1, 2, 3, ...}. This is the first transfinite ordinal — the smallest set that contains all natural numbers. Then ω + 1 = ω ∪ {ω} = {0, 1, 2, ..., ω}, and ω + 2 = {0, 1, 2, ..., ω, ω+1}, and so on. The sequence extends far beyond ω: ω·2, ω², ω^ω, and beyond. Each ordinal is the set of all ordinals before it, which means the collection of all ordinals is well-ordered by the membership relation.
The key structural property that makes ordinals so useful is well-ordering: every nonempty set of ordinals has a least element (its member that contains no other member of the set). You have already encountered the well-ordering principle and mathematical induction for the natural numbers — ordinals generalize exactly this. Transfinite induction works by the same logic: to prove a property holds for all ordinals, show it holds for 0, show that if it holds for all ordinals less than α then it holds for α, and handle the special case of limit ordinals (like ω) that have no immediate predecessor. This three-case structure — zero, successor, limit — is the heartbeat of virtually every transfinite proof.
Ordinals capture *order type*, not just size. Two sets can have the same cardinality but different ordinal structure. The set {0, 1, 2, ...} has order type ω (a first element, no last). The set {..., -2, -1, 0} has no first element — it is not well-ordered — and so has no ordinal. Even among well-orderings of the same size, {1, 2, 3, ..., 0} (natural numbers followed by zero at the end) has order type ω + 1, not ω, because there is now a last element. Ordinals are precisely the canonical representatives of well-ordering types: each well-ordered set is order-isomorphic to exactly one ordinal. This is why ordinals are foundational — they provide the standard yardstick against which any well-ordered structure is measured.